Integrative analysis of large-scale loss-of-function screens identifies robust cancer-associated genetic interactions

Abstract
Introduction
Results
Discussion
Materials and methods
Data availability
References
Article and author information
Metrics

Abstract

Genetic interactions, including synthetic lethal effects, can now be systematically identified in cancer cell lines using high-throughput genetic perturbation screens. Despite this advance, few genetic interactions have been reproduced across multiple studies and many appear highly context-specific. Here, by developing a new computational approach, we identified 220 robust driver-gene associated genetic interactions that can be reproduced across independent experiments and across non-overlapping cell line panels. Analysis of these interactions demonstrated that: (i) oncogene addiction effects are more robust than oncogene-related synthetic lethal effects; and (ii) robust genetic interactions are enriched among gene pairs whose protein products physically interact. Exploiting the latter observation, we used a protein–protein interaction network to identify robust synthetic lethal effects associated with passenger gene alterations and validated two new synthetic lethal effects. Our results suggest that protein–protein interaction networks can be used to prioritise therapeutic targets that will be more robust to tumour heterogeneity.

Introduction

Large-scale tumour genome sequencing efforts have provided us with a compendium of driver genes that are recurrently altered in human cancers (Vogelstein et al., 2013). In some cases, these genetic alterations have been associated with altered sensitivity to targeted therapies. Examples of targeted therapies already in clinical use include approaches that exploit oncogene addictions, such as the increased sensitivity of BRAF mutant melanomas to BRAF inhibitors (Chapman et al., 2011), and approaches that exploit non-oncogene addiction/synthetic lethality, such as the sensitivity of BRCA1/2 mutant ovarian or breast cancers to PARP inhibitors (Lord and Ashworth, 2017). An ongoing challenge is to associate the presence of other driver gene alterations with sensitivity to existing therapeutic agents (Barretina et al., 2012; Iorio et al., 2016) or to identify candidate therapeutic targets whose inhibition may provide therapeutic benefit to patients with specific mutations. Towards this end, multiple groups have performed large-scale loss-of-function genetic perturbation screens in panels of tumour cell lines to identify vulnerabilities that are associated with the presence or absence of specific driver gene mutations (i.e. genetic interactions) (Behan et al., 2019; Campbell et al., 2016; Marcotte et al., 2016; McDonald et al., 2017; Meyers et al., 2017; Tsherniak et al., 2017). Others have performed screens in ‘isogenic’ cell line pairs that differ only by the presence of a specific oncogenic alteration (Martin et al., 2017; Steckel et al., 2012). Despite these large-scale efforts, very few genetic interactions have been identified in more than one study (recently reviewed Ryan et al., 2018). Even in the case of cancer driver genes subjected to multiple screens, such as KRAS, few genetic interactions have been identified in more than one screen (Downward, 2015). This lack of reproducibility may be due to technical issues, for example false positives and false negatives due to inefficient gene targeting reagents (Kaelin, 2012), and/or real biological issues, such as the context specificity of genetic interactions (Henkel et al., 2019; Ryan et al., 2018). We refer to those genetic interactions that can be reproduced across multiple screens and across distinct cell line contexts as robust genetic interactions. Given that tumours exhibit considerable molecular heterogeneity both within and between patients there is a real need to: (i) identify robust genetic interactions that can be reproduced across heterogeneous cell line panels, reasoning that these reproducible effects will be more likely to be robust in the face of the molecular heterogeneity seen in human cancers; (ii) prioritise these robust genetic interactions for further therapeutic development; and (iii) understand the characteristics of robust genetic interactions in cancer as a means to predict new therapeutic targets.

To achieve this, we developed, and describe here, a computational approach that leverages large-scale cell line panel screens to identify those genetic interactions that can be reproducibly identified across multiple independent experiments. We found that for all oncogenes studied, the most significant reproducible dependency identified was an oncogene-addiction rather than a synthetic lethal effect. Excluding oncogene addictions, we found 220 reproducible genetic interactions. In investigating the nature of these robust genetic interactions, we found that they are significantly enriched among gene pairs whose protein products physically interact. This suggests that incorporating prior knowledge of protein–protein interactions may be a useful approach to guide the selection of reproducible ‘hits’ from genetic screens as candidates worth considering as therapeutic targets in cancer. We demonstrate the utility of the approach in identifying robust synthetic lethal interactions from chemogenetic screens and in identifying synthetic lethal interactions associated with ‘passenger’ gene alterations.

Results

A ‘discovery and validation’ approach to the analysis of loss-of-function screens identifies reproducible genetic dependencies

We first wished to identify genetic interactions that could be independently reproduced across multiple distinct loss-of-function screens. To do this, we obtained gene sensitivity scores from four large-scale loss-of-function screens in panels of tumour cell lines, including two shRNA screens (DRIVE [McDonald et al., 2017], DEPMAP [Tsherniak et al., 2017]) and two CRISPR-Cas9 mutagenesis screens (AVANA [Meyers et al., 2017], SCORE [Behan et al., 2019]). We harmonised the cell line names across all studies, so they could be compared with each other and also with genotypic data (Barretina et al., 2012; Iorio et al., 2016; Figure 1A). In total, 917 tumour cell lines were screened in at least one loss-of-function study. Only 50 of these cell lines were common to all four studies while 407 cell lines were included in only a single study (Figure 1B). It is the partially overlapping nature of the screens that motivated the subsequent approach we took for our analysis. We used a ‘discovery set’ and ‘validation set’ approach to identifying genetic interactions across multiple screens - first identifying associations between driver gene alterations and gene inhibition sensitivity in the discovery study and then testing the discovered dependency in the validation study (Figure 1C). However, to ensure that any reproducibility observed was not merely due to cell lines common to both datasets, we first removed cell lines from the validation dataset if they were present in the discovery dataset (Figure 1C). For example, when using DEPMAP as the discovery dataset and AVANA as the validation dataset, we performed the validation analysis on the subset of cell lines that were present in AVANA but not in DEPMAP. In doing so, we ensured that any genetic interactions discovered were reproducible across different screening platforms (either distinct gene inhibition approaches, that is shRNA vs CRISPR, or distinct shRNA/CRISPR libraries) and also robust to the molecular heterogeneity seen across different cell line panels.

Figure 1

Download asset Open asset

Identifying robust genetic interactions using partially overlapping loss-of-function screens.

(A) Workflow showing the integration of four different loss-of-function screen datasets. (B) Venn diagram showing the overlap of cell lines between the four datasets analysed in this study. (C) Workflow showing how robust genetic interactions are identified using discovery and validation sets.

Similar to our previous work (Bridgett et al., 2017; Campbell et al., 2016) we integrated copy number profiles and exome sequencing data to annotate all cell lines according to whether or not they featured likely functional alterations in any one of a panel of cancer driver genes (Vogelstein et al., 2013) (see Materials and methods, Supplementary file 1). We then identified associations between driver gene alterations and sensitivity to the inhibition of specific genes using a multiple regression model. This model included tissue type, microsatellite instability and driver gene status as independent variables and gene sensitivity score as the dependent variable (Materials and methods). Microsatellite instability was included as a covariate as it has previously been shown to be associated with non-driver gene specific dependencies (Behan et al., 2019), while tissue type was included to avoid confounding by tissue type. We focused this analysis on ‘selectively lethal’ genes - that is those genes whose inhibition killed some, but not all cell lines (Materials and methods, Supplementary file 2). We analysed each pair of screens in turn and considered a genetic dependency to be reproducible if it was validated in at least one discovery/validation pair. Using this approach, we tested 142,477 potential genetic dependencies between 61 driver genes and 2421 selectively lethal genes. We identified 1530 dependencies that were significant in at least one discovery screen (Figure 2A, Figure 2—figure supplement 1). All 61 driver genes had at least one dependency that was significant in at least one discovery screen while less than half of the selectively lethal genes (1,141/2,421) had a significant association with a driver gene. Of the 1530 dependencies that were significant in at least one discovery screen, only 229 could be validated in a second screen (Supplementary file 3, Figure 2A). For example, in the AVANA dataset TP53 mutation was associated with resistance to inhibition of both MDM4 and CENPF, but only the association with MDM4 could be validated in a second dataset (Figure 2B and C). Similarly, in the DEPMAP dataset NRAS mutation was associated with increased sensitivity to the inhibition of both NRAS itself and ERP44, but only the sensitivity to inhibition of NRAS could be validated in a second dataset (Figure 2B and C).

Figure 2 with 1 supplement see all

Download asset Open asset

Discovered and validated genetic dependencies.

(A) Scatterplot showing the genetic dependencies identified across all datasets. Each individual point represents a gene pair, the x-axis shows the common language effect size, and the y-axis shows the -log10 p-value from the discovery dataset. Selected gene pairs are highlighted – the driver gene is listed first, followed by the associated dependency. Each gene pair may have been tested in multiple discovery studies, only the interaction with the most significant discovery p-value is shown. Scatterplots for individual studies are presented in Figure 2—figure supplement 1. (B) Tukey boxplots showing examples of robust genetic dependencies, including an increased resistance of *TP53* mutant tumour cell lines to *MDM4* inhibition and increased sensitivity of *NRAS* mutant tumour cell lines to *NRAS* inhibition. In each box plot the top and bottom of the box represents the third and first quartiles and the box band represents the median; whiskers extend to 1.5 times the interquartile distance from the box. WT = wild type, ALT = altered. Throughout blue is used to indicate increased sensitivity (synthetic lethality or oncogene addiction), yellow to indicate resistance to inhibition of the target gene. (C) Boxplots showing examples of genetic dependencies discovered but not validated, including an increased resistance of *TP53* mutant cell lines to *CENPF* inhibition and increased sensitivity of *NRAS* mutant cell lines to *ERP44* inhibition.

The 229 reproducible dependencies involved 31 driver genes and 204 selectively lethal genes. Of the 229 reproducible genetic dependencies nine were ‘self vs. self’ associations, where the alteration of a gene was associated with sensitivity to its own inhibition. The majority (7/9) of these ‘self vs. self’ associations were oncogene addiction effects, such as the increased sensitivity of NRAS mutant cell lines to NRAS inhibition (Figure 2B). Similarly, we identified robust oncogene addictions involving the CTNNB1 (β-Catenin), KRAS, EGFR, BRAF, ERBB2 and PIK3CA oncogenes (Figure 3A and B, Figure 3—figure supplement 1B). For EGFR and CTNNB1, the only identified robust dependency was an oncogene addiction effect. For all other oncogenes there were additional robust dependencies identified, but in all cases the most significant reproducible dependency was an oncogene addiction (Figure 3—figure supplement 1A). These observations suggest that for most oncogenes the oncogene addiction effect might be more robust than any oncogene-related synthetic lethal effects.

Figure 3 with 1 supplement see all

Download asset Open asset

Identified robust genetic interactions.

(A) Dot plot showing the robust genetic dependencies identified for oncogenes. Each coloured circle indicates a robust genetic dependency, scaled according to the number of dataset pairs it was validated in. The most significant genetic dependency (lowest FDR in a validation set) for each driver gene is labelled. Oncogenes are sorted by the number of robust dependencies and the total number of robust genetic dependencies for each driver gene is shown in parentheses. (B) Example of a validated oncogene addiction – *ERBB2* amplified cells are sensitive to *ERBB2* inhibition. Left shows the discovery dataset (SCORE) and right shows the validation dataset (AVANA). (C) Dot plot showing the robust genetic interactions identified for all driver genes. Each coloured circle indicates a robust genetic interaction, scaled according to the number of dataset pairs it was validated in. The most significant genetic interaction (lowest FDR in a validation set) for each driver gene is labelled. Drivers are sorted by the number of robust interactions and the total number of robust genetic interactions for each driver gene is shown in parentheses. *TP53* (132 robust genetic interactions) has been excluded for clarity, as have all self-self dependencies. (D) Examples of known genetic interactions identified from the integrated analysis, including an increased sensitivity of *PTEN* mutant tumour cell lines to *PIK3CB* inhibition and increased resistance of *TP53* mutant tumour cell lines to *MDM2* inhibition. Top row shows the data used to discover the interactions (DEPMAP dataset) while the bottom row shows the data used to validate the interactions (AVANA dataset with cell lines from DEPMAP excluded). (E) Synthetic lethal interactions involving paralog pairs. Top row shows the data used to discover the interactions (DRIVE dataset) while the bottom row shows the data used to validate the interactions (AVANA dataset with cell lines from DRIVE excluded).

We also identified two (2/9) examples of ‘self vs. self’ dependencies involving tumour suppressors -TP53 (aka p53) and CDKN2A (aka p16/p14arf) (Figure 3—figure supplement 1C). This type of relationship has previously been reported for TP53: TP53 inhibition appears to offer a growth advantage to TP53 wild type cells but not to TP53 mutant cells (Giacomelli et al., 2018). Inhibiting TP53 in TP53 mutant cells has a largely neutral effect, while on average inhibiting TP53 in TP53 wild type cells actually increases fitness growth. Consequently, we observed an association between TP53 status and sensitivity to TP53 inhibition. Similar effects were seen for CDKN2A, although the growth increase resulting from inhibiting CDKN2A in wild-type cells is much lower than that seen for TP53 (Figure 3—figure supplement 1C). These ‘self vs. self’ dependencies, in particular the oncogene addictions, serve as evidence that our approach could identify well characterised genetic associations. However, as our primary interest was in genetic interactions between different genes, we excluded ‘self vs. self’ interactions from further analysis, leaving us with 220 robust genetic interactions (Figure 3C).

Many robust genetic interactions reflect known pathway structure

Many of the robust genetic interactions we identified have been previously reported, including both sensitivity relationships, such as increased sensitivity of PTEN mutant cell lines to inhibition of the phosphoinositide 3-kinase-coding gene PIK3CB (Wee et al., 2008), and resistance relationships, such as an increased resistance of TP53 mutant cell lines to MDM2 inhibition (Figure 3D).

Amongst the set of 220 robust genetic interactions, we identified two previously reported ‘paralog lethalities’ – synthetic lethal relationships between duplicate genes (paralogs) (Helming et al., 2014; Hoffman et al., 2014; Oike et al., 2013; Figure 3E). We found a robust association between mutation of the tumour suppressor ARID1A and sensitivity to inhibition of its paralog ARID1B (Helming et al., 2014) and also an association between mutation of SMARCA4 and sensitivity to inhibition of its paralog SMARCA2 (Hoffman et al., 2014; Oike et al., 2013). Both pairs of genes (ARID1A/ARID1B and SMARCA4/SMARCA2) encode components of the larger SWI/SNF complex (Wilson and Roberts, 2011).

Some of the robust genetic dependencies could be readily interpreted using known pathway structures. For instance, many of the robust dependencies associated with the oncogene BRAF could be interpreted in terms of BRAF’s role in the MAPK pathway. BRAF mutation was associated with increased sensitivity to inhibition of its downstream effectors MEK (MAP2K1) and ERK (MAPK1), and increased resistance to inhibition of the alternative RAF isoform gene CRAF (RAF1) and the MAPK regulators PTPN11 and SHOC2 (Figure 4A and B; Hill et al., 2019). BRAF mutation was also associated with increased sensitivity to inhibition of PEA15, presumably a result of the requirement of PEA15 for ERK dimerisation and signalling activity (Formstecher et al., 2001; Herrero et al., 2015).

Figure 4

Download asset Open asset

Robust genetic interactions involving *RB1* and *BRAF* recapitulate pathway relationships.

(A) Simplified RAS/RAF/MEK/ERK pathway diagram. Protein names (e.g. MEK) are shown inside nodes, while associated gene names are shown adjacent (e.g. *MAP2K1*). Nodes are coloured according to their association with BRAF mutation - blue indicates increased sensitivity of *BRAF* mutant cell lines, yellow indicates increased resistance. (B) Boxplots showing selected genetic interactions associated with *BRAF* mutation. (C) Simplified Rb pathway diagram, highlighting robust genetic interactions involved in the Rb pathway. (D) Boxplots showing selected genetic interactions associated with *RB1* alteration.

Mutation or deletion of the tumour suppressor RB1 (Retinoblastoma 1, Rb) was associated with increased sensitivity or resistance to inhibition of multiple Rb pathway members (Figure 4C and D). We found that RB1 loss was reproducibly associated with resistance to inhibition of its negative regulators CDK4 and CDK6, consistent both with the known Rb pathway structure and with preclinical data suggesting that RB1 mutation confers resistance to CDK4/6 inhibitors (Asghar et al., 2015; O'Leary et al., 2018). Rb is a negative regulator of multiple E2F transcription factors, and we found that RB1 loss was reproducibly associated with increased sensitivity to both E2F1 and E2F3 inhibition (Figure 4C and D). RB1 loss was also associated with robust sensitivity to SKP2, a binding partner of Rb (Ji et al., 2004) first identified as an RB1 synthetic lethal partner in retinoblastoma (Xu et al., 2014) and more recently as a highly penetrant RB1 synthetic lethal partner in triple negative breast cancer (Brough et al., 2018; Figure 4C and D). Finally, RB1 loss was reproducibly associated with increased sensitivity to inhibition of Cyclin Dependent Kinase 2 (CDK2), suggesting that it may be a useful biomarker for CDK2-specific inhibitors (Tadesse et al., 2019).

To test if this enrichment of pathway members among the robust dependencies associated with specific driver genes was a common phenomenon, for each driver gene with at least three dependencies we asked if these dependencies were enriched in specific signalling pathways (see Materials and methods). Of the twelve driver genes tested, we found that five of these were enriched in specific pathways and in all five cases found that the driver gene itself was also annotated as a member of the most enriched pathway (Supplementary file 4). As expected RB1 (most enriched pathway ‘G1 Phase’) and BRAF (most enriched pathway ‘Negative feedback regulation of MAPK pathway’) were among the five driver genes, alongside PTEN (‘PI3K/AKT activation’), CDKN2A (‘Cell cycle’), and NRAS (‘Ras signaling pathway’).

Robust genetic interactions are enriched in protein–protein interaction pairs

In seeking to understand what particular characteristics robust genetic interactions might have, we noted that many of the robust genetic interactions we identified involved gene pairs whose protein products operate in the same pathway (e.g. the Rb pathway) or protein complex (e.g. SWI/SNF) suggesting that genetic interactions between gene pairs whose protein products physically interact may be more robust than other genetic interactions. To test this hypothesis, we compared the robust genetic interactions we identified with protein–protein interactions from the STRING protein–protein interaction database (Szklarczyk et al., 2015). We found that, when considering the set of all gene pairs tested, gene pairs identified as significant genetic interactors in at least one dataset are more likely to encode proteins that physically interact (Figure 5A) (Odds Ratio (OR) = 4.0, p<2×10⁻¹⁶, Fisher’s Exact test). Furthermore, of the genetic interactions identified as significant in at least one dataset, those that are supported by a protein–protein interaction were significantly more likely to be reproduced in a second dataset (Figure 5A) (OR = 3.9, p<1×10⁻¹³). We therefore concluded that protein–protein interaction pairs are more likely to be significant hits in one dataset and even more likely to be reproduced across multiple datasets, suggesting this might be a feature of robust genetic interaction effects.

Figure 5 with 2 supplements see all

Download asset Open asset

Robust genetic interactions are enriched in protein–protein interaction pairs.

(A) Barchart showing the percentage of protein–protein interacting pairs observed among different groups of gene pairs. The groups represent all gene pairs tested, gene pairs found to be significantly interacting in at least one screen (FDR < 20%), and gene pairs found to reproducibly interact across multiple screens (i.e. a discovery and validation screen). Stars (*) indicate significant differences between groups, all significant at p<0.001 using Fisher’s Exact Test. Odds ratios and p-values are provided in Supplementary file 5. (B) As A but with interactions associated with *TP53* removed. (C) As B but here the discovery and validation sets contain the same cell lines screened in different studies (e.g. ‘AVANA ∩ DEPMAP’ as discovery and ‘DEPMAP ∩ AVANA’ as validation). Consequently, reproducibility here means ‘technical reproducibility’ using different screening platforms. (D) Similar to B but here the discovery and validation sets contain single datasets partitioned into non-overlapping cell line sets (e.g. ‘AVANA \ DEPMAP’ as discovery and ‘AVANA ∩ DEPMAP’ as validation). Consequently, reproducibility here means ‘genetic robustness’ - the same association between gene pairs is observed across distinct genetic backgrounds.

We noted that a large number (n = 132) of robust genetic interactions involved TP53, presumably as a result of the high number of TP53 mutant tumour cell lines in the datasets (and its high mutation frequency in human cancer) and the associated increased statistical power to detect TP53-related genetic interactions. We therefore considered whether the significant number of TP53-related genetic interactions in our dataset could confound our analyses, especially as TP53 is also associated with a disproportionately high number of protein interactions (>1700 medium confidence interactors in the STRING database alone, compared to a median of 37 medium confidence interactions across all proteins). However, even after excluding genetic interactions involving TP53, the observation that robust genetic interactions were enriched in protein–protein interaction pairs was still evident (Figure 5B); known protein interaction pairs were more likely to be identified as significant genetic interactions in one screen (OR = 3.8, p<2×10⁻¹⁶) and among the significant genetic interactions discovered in one screen, those involving protein–protein interaction pairs were more likely to be reproduced in a validation screen (OR = 9.3, p<2×10⁻¹⁶). The same effects were observed when considering genetic interactions observed at different false-discovery rate (FDR) thresholds (Figure 5—figure supplement 1A,B) and using different sources of protein–protein interaction data (Figure 5—figure supplement 1C,D,E, Supplementary file 5; Alanis-Lobato et al., 2017; Chatr-Aryamontri et al., 2015; Huttlin et al., 2020).

Previous work has demonstrated that the protein–protein interaction networks aggregated in databases are subject to significant ascertainment bias – some genes are more widely studied than others and this can result in them having more reported protein–protein interaction partners than other genes (Rolland et al., 2014). As cancer driver genes are studied more widely than most genes, they may be especially subject to this bias. To ensure the observed enrichment of protein–protein interactions among genetically interacting pairs was not simply due to this ascertainment bias, we compared the results observed for the real STRING protein–protein interaction network with 100 degree-matched randomised networks and again found that there was a higher than expected overlap between protein–protein interactions and both discovered and validated genetic interactions (Figure 5—figure supplement 2).

The increased reproducibility of genetic interactions associated with protein–protein interactions across different genetic perturbation screen datasets could have two distinct causes - increased reproducibility across distinct technologies or libraries (e.g. CRISPR/shRNA) or increased reproducibility/robustness of genetic interactions in cell line panels with distinct molecular backgrounds. To test the former possibility, we repeated our discovery/validation approach but focused on the set of cell lines that were common to different genetic perturbation screen datasets. Using this approach, the molecular backgrounds (i.e. cell lines) tested were the same, but the screening approach or library used differed. Upon doing this, we found that genetic interactions between gene pairs whose protein products physically interact were significantly more reproducible across studies (Figure 5C, OR = 6.1 and p<2×10⁻¹⁰ when compared to discovered genetic interactions) (Supplementary file 5). To test reproducibility using the same screening approach across molecularly distinct cell lines, we artificially split individual datasets into non-overlapping discovery and validation sets of cell lines. Again, we found that genetic interactions between gene pairs whose protein products physically interact were more reproducible across distinct cell line panels (Figure 5D, OR = 8.0 and p<1×10⁻¹² when compared to discovered genetic interactions) (Supplementary file 5). We therefore concluded that genetic interactions supported by protein–protein interactions were more reproducible across different screening approaches and across distinct cell line contexts, suggesting that these interactions are, overall, more robust.

Prioritising robust synthetic lethal interactions from chemogenetic screens

As an alternative to genetic perturbation screening in large cell line panels, genetic interactions can also be identified using chemogenetic screens, where loss-of-function screens are performed in the presence and absence of specific small molecule inhibitors whose targets are relatively well defined. Based on the observations made earlier, we hypothesised that genetic interactions identified in chemogenetic screens that involved genes whose protein products physically interact with the target of the inhibitor should both be more likely to be identified as genetic interaction partners in one screen and also more likely to be reproduced across multiple screens (i.e. to be more robust). To test this hypothesis, we analysed the results of a recent chemogenetic screen performed to identify genes whose loss is synthetic lethal with ATR inhibition (Wang et al., 2019). In this study, genome-wide CRISPR-Cas9 screens in three cell lines from different histologies (breast, kidney, colon) were used to identify genes whose inhibition is selectively essential in the presence of a small molecule ATR kinase inhibitor (Vendetti et al., 2015; Wang et al., 2019; Figure 6A).

Figure 6

Download asset Open asset

Reproducible ATR synthetic lethal interactions are enriched in ATR protein–protein interaction partners.

(A) Workflow - synthetic lethal interactions from CRISPR-Cas9 screens in three cell lines (Wang et al., 2019) were compared to identify reproducible synthetic lethal partners. These genes were then compared with known ATR protein–protein interaction partners from the STRING database. (B) Bar chart showing the percentage of ATR protein interaction partners observed in different groups of genes. Genes are grouped according to whether they were identified as an ATR synthetic lethal partner in 0, 1, 2, or 3 cell line screens. Comparisons between all pairs of groups are significant at p<0.001 (Fisher’s exact test) except for the comparison between genes that were hits in 2 and 3 cell lines (p=0.06).

As predicted, we found that protein interaction partners of ATR are more likely than random genes to be identified as a significant synthetic lethal interactor of ATR in at least one cell line (Figure 6B). Furthermore, we found that among the synthetic lethal interactions identified in at least one cell line, those involving known ATR protein interaction partners were significantly more likely to be reproduced in a second or even third cell line (Figure 6B). This suggests that, of the candidate genes identified in one screen, those that encode protein–protein interaction partners of ATR are significantly more likely to validate in additional contexts than genes with no known functional relationship to ATR.

Prioritising robust synthetic lethal interactions involving passenger gene alterations

In addition to alterations of cancer driver genes, tumour cells typically harbour genetic alterations of large numbers of ‘passenger’ genes. Although these genes may not facilitate tumourigenesis or promote cancer cell growth, their alteration may still impart genetic vulnerabilities upon tumour cells. Indeed, multiple synthetic lethal interactions have been identified involving passenger genes that exhibit recurrent copy number loss in cancer cells due to their chromosomal proximity to tumour suppressor genes lost via loss-of-heterozygosity (Kryukov et al., 2016; Marjon et al., 2016; Mavrakis et al., 2016; Muller et al., 2012; Muller et al., 2015). The space of genetic interactions to test involving passenger gene alterations is much larger than that involving driver genes, as nearly every gene in the genome is either mutated or deleted in some cancer context. In addition, passenger genes are typically altered at frequencies lower than for driver genes and therefore the statistical power to identify genetic interactions associated with their alteration is somewhat reduced. With these issues in mind, we reasoned that protein–protein interaction maps might help narrow the search space considerably and thus reduce the burden of multiple hypothesis testing. For all passenger genes that were recurrently lost in at least ten tumour cell lines, either through homozygous deletion or loss-of-function mutation, we searched for genetic interactions with their protein–protein interaction partners using the same discovery and validation approach previously used for driver genes. In total we tested 47,781 interactions involving 2,972 passenger genes and 2149 selectively lethal genes. To perform an all-against-all test without filtering based on protein–protein interactions would have required more than six million tests, significantly increasing the burden of multiple-hypothesis testing. At an FDR of 20% we found 11 robust genetic interactions involving passenger gene alterations (Supplementary file 7). To assess whether this is more than would be expected by chance we randomly sampled 47,781 gene pairs from the same search space 100 times. The median number of robust genetic interactions identified amongst these randomly sampled gene pairs was 1 (mean 1.27, min 0, max 6) suggesting that the 11 robust genetic interactions observed among protein–protein interacting pairs was more than would be expected by chance. Three of the robust interactions involve genes frequently deleted with the tumour suppressor CDKN2A (CDKN2B and MTAP) and recapitulate known associations with CDKN2A. A further two genetic interactions involve a single chromosomal region (19p21.3) containing two interferon genes (IFNB1 and IFNW1) which are frequently deleted together and consequently these two interactions really represent a single association (an increased sensitivity to thrombopoietin receptor MPL). Of the six remaining genetic interactions identified, four represent examples of paralog lethalities – loss of one member of a paralog pair is associated with increased sensitivity to the inhibition of the other member. RPL22 loss was associated with increased sensitivity to its paralog RPL22L1, TIMM17B with its paralog TIMM17A, DDX17 with its paralog DDX5, and VPS4B with its paralog VPS4A. We selected two of these robust synthetic lethal interactions for further validation – VPS4B/VPS4A and DDX17/DDX5.

VPS4A and VPS4B are highly sequence similar whole genome duplicates with protein sequence identity of 81%. Both proteins can form a complex with the Vacuolar protein sorting-associated protein VTA1 (Huttlin et al., 2017) and are involved in endosomal trafficking. VPS4B is located at 18q21 and is frequently deleted with the tumour suppressor SMAD4, explaining the relatively high frequency of loss of VPS4B in cancer. Previous analysis of the DRIVE shRNA dataset identified an association between VPS4B copy number loss and VPS4A sensitivity (McDonald et al., 2017). Here we find evidence of this association in two additional datasets – AVANA and SCORE (Figure 7A). Although this association is robust, it does not establish a causal link between VPS4B loss and VPS4A sensitivity. Indeed, there are 39 protein-coding genes on chromosome 18 located between SMAD4 and VPS4B, any one of which could cause sensitivity to VPS4A inhibition. To verify that VPS4B is the cause of VPS4A sensitivity we transfected isogenic knockouts of either VPS4A and VPS4B with siRNA smartpools targeting either VPS4A or VPS4B and found that, consistent with a negative genetic interaction between the two genes, compared to wildtype parental cells VPS4A knockout cells were more sensitive to siRNA targeting VPS4B and VPS4B knockout cells were more sensitive to siRNA targeting VPS4A (Figure 7B, Supplementary file 8).

Figure 7

Download asset Open asset

Robust synthetic lethalities associated with passenger gene loss.

(A) Boxplots showing the association between *VPS4B* loss and *VPS4A* sensitivity in the discovery dataset (DRIVE) and two validation datasets (AVANA and SCORE). (B) Mean viability of HAP1 cells treated with siRNA smartpools targeting *VPS4A* or *VPS4B*. Individual data points are shown as black dots. Data are normalized within each cell line such that the mean viability of cells treated with a negative control (non-targeting scrambled siRNA) is equal to one and the mean viability treated with a positive control (siRNA smartpool targeting the broadly essential *PLK1* gene) is equal to 0. P-values from two-sided heteroscedastic T-tests. (C) Boxplots showing the association between *DDX17* loss and *DDX5* sensitivity in the discovery dataset (AVANA) and the validation dataset (DRIVE). (D) Mean viability of HAP1 cells treated with siRNA smartpools targeting *DDX5* or *DDX17*, normalization and statistics as per B.

Like VPS4A and VPS4B, DDX5 and DDX17 are widely conserved highly sequence similar whole genome duplicates (protein sequence identity 69%). They are DEAD box family RNA helicases that have multiple roles in both transcription and splicing; they act as coregulators for multiple transcription factors and also function as components of the spliceosome (Dardenne et al., 2014; Fuller-Pace, 2013). A direct protein–protein interaction between the two genes has also been reported (Hegele et al., 2012; Huttlin et al., 2017). DDX17 is located at 22q12 in close proximity to the tumour suppressor MYH9, potentially explaining its recurrent deletion in tumour cell lines. We identified an association between DDX17 loss and DDX5 sensitivity in the AVANA CRISPR dataset and validated this association in the DRIVE shRNA dataset (Figure 7C). As with VPS4A/VPS4B, to verify that DDX17 loss is the cause of DDX5 sensitivity we transfected isogenic knockouts of either DDX17 and DDX5 with siRNA smartpools targeting either DDX5 or DDX17. We found that, consistent with a negative genetic interaction between the two genes, compared to wildtype parental cells DDX17 knockout cells were more sensitive to siRNA targeting DDX5 and DDX5 knockout cells were more sensitive to siRNA targeting DDX17 (Figure 7D, Supplementary file 8).

Discussion

While the reproducibility of pharmacogenomic screens in cancer cell lines has been much discussed (Cancer Cell Line Encyclopedia Consortium and Genomics of Drug Sensitivity in Cancer Consortium, 2015; Haibe-Kains et al., 2013; Niepel et al., 2019), relatively little attention has been paid to the reproducibility of results from large-scale genetic screens in cell lines. Analyses of the pharmacogenomic screen datasets have primarily focused on reproducibility in a very strict sense - that is quantifying the extent to which the same drug elicits the same response in the same cell line when assayed across different sites (Niepel et al., 2019). In some cases these analyses have been extended to quantify the extent to which the same associations between biomarkers and drugs can be observed across the same cell line panels assayed in different experiments (Cancer Cell Line Encyclopedia Consortium and Genomics of Drug Sensitivity in Cancer Consortium, 2015). Here we were interested in reproducibility in a much broader sense and sought to identify genetic interactions that could be reproduced both across distinct experiments and across distinct cell line panels, that is interactions that are robust to genetic and molecular heterogeneity. We developed an approach to identify these robust genetic interactions and used it to identify a set of 220 robust genetic interactions associated with cancer driver genes. We found that these robust genetic interactions are enriched among gene pairs whose protein products physically interact, suggesting a means by which we might prioritise the most promising candidates from screens for follow on studies.

We do not claim that our set of robust genetic interactions is comprehensive, as there are many reasons why real robust genetic interactions may not be identified by our approach. There are many driver genes that we have not included in our analysis because they are infrequently mutated in the datasets studied. Consequently, we can report no interactions for these genes. We have also focussed only on identifying interactions associated with mutation or copy number changes. There are likely to be dependencies associated with altered gene/protein expression that will be missed by this approach. Furthermore, for the genes that we do analyse, it is likely that some real interactions are not detected due to a lack of statistical power. Finally, of the dependencies identified in a discovery screen but absent in a validation screen, false negatives due to reagents with poor gene targeting ability likely play a significant role (Kaelin, 2012).

We have exclusively focussed on identifying dependencies that are evident across panels of cell lines from multiple cancer types (‘pan-cancer dependencies’). It is likely that there are robust dependencies only evident within specific cancer types, but it is difficult to use our approach to identify them due to the restricted number of cell lines available for each cancer type. Even with a relatively common mutation (e.g. KRAS mutation in non-small cell lung cancer) it is challenging to partition the available cell lines into distinct discovery and validation sets while maintaining statistical power to identify potential dependencies. This issue may be alleviated by efforts to create large numbers of new tumour cell lines (Boehm and Golub, 2015) or through using isogenic models for discovery and cell line panels for validation (Ryan et al., 2018).

Many published synthetic lethal screens have focussed on identifying new drug targets for ‘undruggable’ oncogenes such as MYC and RAS (reviewed in Cermelli et al., 2014 and Downward, 2015 respectively). The rationale for such studies is that the oncogene addiction itself cannot be exploited directly and consequently a synthetic lethal approach is needed. However, here we found that for all oncogenes studied the most significant reproducible dependency identified was an oncogene addiction (Figure 3A). This suggests that any synthetic lethal interactions that are identified for oncogenes will likely be of a smaller effect size or operate in a more restricted context than the oncogene addiction itself. Previous work has suggested this to be true of KRAS (McDonald et al., 2017) but here we find that it appears to be a general property of all oncogenes studied. This implies that wherever possible, direct targeting of oncogenes might be more therapeutically effective than exploiting oncogene-related synthetic lethal effects.

Our approach to identify robust genetic dependencies involving cancer driver genes is unbiased in the sense that we did not incorporate prior knowledge of functional relationships to identify candidate gene pairs to test. Nonetheless, many of the robust synthetic lethalities identified reflect known biology. In particular, for each of the well-studied tumour suppressors ARID1A, SMARCA4 and PTEN the most significant robust synthetic lethal interaction we identified has previously been reported in the literature. For ARID1A, its known synthetic lethal partner ARID1B was the only robust candidate interaction identified while for PTEN and SMARCA4 their established synthetic lethal partners (PIK3CB and SMARCA2 respectively) are the most significant robust hits by a large margin (Figure 3C). As with oncogenes, this suggests that if novel single gene vulnerabilities for these drivers are to be discovered, they may have a smaller effect size or operate in a more restricted setting.

In budding and fission yeast, multiple studies have shown that genetic interactions are enriched among protein–protein interaction pairs and vice-versa (Costanzo et al., 2010; Kelley and Ideker, 2005; Michaut et al., 2011; Roguev et al., 2008). Pairwise genetic interaction screens in individual mammalian cell lines have also revealed an enrichment of genetic interactions among protein–protein interaction pairs (Han et al., 2017; Roguev et al., 2013). Our observation that discovered genetic interactions are enriched in protein–protein interaction pairs is consistent with these studies. However, these studies have not revealed what factors influence the conservation of genetic interactions across distinct genetic backgrounds, that is what predicts the robustness of a genetic interaction. In yeast, the genetic interaction mapping approach relies on mating gene deletion mutants and consequently the vast majority of reported genetic interactions are observed in a single genetic background (Tong et al., 2001). In mammalian cells, pairwise genetic interaction screens across multiple cell lines have revealed differences across cell lines but not identified what factors influence the conservation of genetic interactions across cell lines (Shen et al., 2017). While variation of genetic interactions across different strains or different genetic backgrounds has been poorly studied, previous work has analysed the conservation of genetic interactions across species and shown that genetic interactions between gene pairs whose protein products physically interact are more highly conserved (Roguev et al., 2008; Ryan et al., 2012; Srivas et al., 2016). Our analysis here suggests that the same principles may be used to identify genetic interactions conserved across genetically heterogeneous tumour cell lines. Previous work has also shown that genetic interactions between gene pairs involved in the same biological process, as indicated by annotation to the same gene ontology term, are more highly conserved across species (Ryan et al., 2012; Srivas et al., 2016). Similarly, genetic interactions that are stable across experimental conditions (e.g. in the presence and absence of different DNA damaging agents) are more likely to be conserved across species (Srivas et al., 2016). Although we have not tested them here, these additional features predictive of between-species conservation may also be predictive of robustness to genetic heterogeneity. Our set of robust genetic interactions may serve as the starting point for such analyses and may also serve as a training set for computational approaches to predict synthetic lethality (Jerby-Arnon et al., 2014).

Our finding that the robust genetic interactions associated with some driver genes can be interpreted in terms of the signalling pathway that the driver gene functions in suggests that pathway structure may also provide information on robustness. For example, it seems reasonable to hypothesise that synthetic lethal proteins that are in close vicinity in a pathway are more likely to exhibit a robust synthetic lethal than those that are more distantly connected. However, to allow such a hypothesis to be tested, we believe the annotation of molecular pathways should be somewhat more reliable and the set of experimentally-validated robust genetic interactions much larger.

Our results suggest that knowledge of protein–protein interactions could be used to improve the design and analysis of loss-of-function screens for synthetic lethal interactions. We have demonstrated the utility of incorporating such prior knowledge for identifying robust synthetic lethal interactions from genome-wide chemogenetic screens. We have also demonstrated that protein–protein interactions can aid the identification of genetic interactions associated with passenger gene alterations, where statistical power is limited due to relatively infrequent alterations and the number of potential interactions to test is enormous. An alternative to these approaches, where knowledge of protein–protein interactions is used after the screen has already been performed, would be to screen target libraries for specific driver genes based on their known protein interaction partners. Regardless of the approach used to identify candidate synthetic lethal interactions in a large-scale screen, our results suggest that candidates supported by a protein–protein interaction should be prioritised for follow on study as they are more likely to be robust to the genetic heterogeneity observed in tumour cells.

Materials and methods

All data analysis was performed using Python 3.7 (RRID:SCR_008394), Pandas 0.24 (McKinney, 2011) (RRID:SCR_018214) and StatsModels 0.9.0 (Seabold and Perktold, 2010) (RRID:SCR_016074).

Loss of function screens

Request a detailed protocol

Different scoring systems have been developed for calculating ‘gene level’ sensitivity scores from loss-of-function screens performed with multiple gene targeting reagents per gene (i.e. shRNAs or gRNAs). For the analysis of all loss-of-function screens we used the original authors’ own preferred approaches. CERES sensitivity scores (Meyers et al., 2017) for AVANA (release 18Q4) were obtained from the DepMap portal (https://depmap.org/portal/download/) (RRID:SCR_017655), while DEMETER v2 gene sensitivity scores for the DEPMAP shRNA screen (Tsherniak et al., 2017) were obtained from the same resource. For the DEPMAP screen, some genes were only screened in a subset of cell lines and these were excluded from all analyses. Quantile normalized CRISPRcleaned (Iorio et al., 2018) depletion log fold changes for Project SCORE (Behan et al., 2019) were obtained from the Project SCORE database (https://score.depmap.sanger.ac.uk/). ATARIS (Shao et al., 2013) scores for the DRIVE dataset (McDonald et al., 2017) were obtained from the authors. 28 of the 398 cell lines screened in DRIVE had missing gene scores for ~25% of genes screened and these cell lines were excluded from further analysis. All screens were mapped to a common cell line name format (that followed by the Cancer Cell Line Encyclopaedia [Barretina et al., 2012]) using the Cell Model Passports resource where appropriate (van der Meer et al., 2019).

Identifying selectively lethal genes

Request a detailed protocol

Similar to previous work (McDonald et al., 2017; Tsherniak et al., 2017), to reduce the burden of multiple hypothesis testing we focused our analysis on genes whose inhibition appeared to cause growth defects in subsets of the cancer cell lines screened. That is, rather than testing for associations with genes whose inhibition was always lethal or never lethal, we focused our analyses on genes that could be associated with distinct sensitive and resistant cell line cohorts. We first identified a set of ‘selectively lethal’ genes using the AVANA dataset (Meyers et al., 2017) - those with a gene sensitivity score <−0.6 in at least 10 cell lines but no more than 259 cell lines (half of the screened cell lines). We augmented this with a list of 65 ‘outlier genes’ identified by the authors of the DRIVE study as having a skewed distribution suggesting distinct sensitive and resistant cohorts (McDonald et al., 2017). Finally from the combined list we removed genes known to be commonly essential in cancer cell lines (Hart et al., 2017). This resulted in a set of 2470 selectively lethal genes (Supplementary file 2) which were used for all association analyses.

Identifying gene alterations from copy number and exome profiling

Request a detailed protocol

For all cell lines we obtained sequencing data (CCLE_DepMap_18q3_maf_20180718.txt) and copy number profiles (public_18Q3_gene_cn_v2.csv) from the DepMap portal. These datasets contain integrated genotyping data from both the Cancer Cell Line Encyclopedia and GDSC resources (Barretina et al., 2012; Cancer Cell Line Encyclopedia Consortium and Genomics of Drug Sensitivity in Cancer Consortium, 2015; Iorio et al., 2016). We used this to identify likely functional alterations in a panel of cancer driver genes (Vogelstein et al., 2013) restricting our analysis to those genes that were subject to targeted sequencing as part of the Cancer Cell Line Encyclopedia (Barretina et al., 2012).

For most oncogenes we considered the gene to be functionally altered if it contained a protein altering mutation at a residue that is recurrently altered in either the COSMIC database or the Cancer Genome Atlas. For a small number of oncogenes (ERBB2, CCND1, MDM2, MDM4) we considered them to be functionally altered only if they were amplified. For all tumour suppressors we considered all protein-coding mutations and homozygous deletions to be functional alterations. We used a threshold of −1.28 to call homozygous deletions. The matrix of functional alterations to driver genes is presented in Supplementary file 1.

To identify loss-of-function alterations to passenger genes, a similar pipeline was used. However the driver genes from Vogelstein et al., 2013 were excluded from analysis and only clear loss-of-function alterations were considered to be functional. The matrix of gene loss in passenger genes is presented in Supplementary file 6.

Identifying genetic dependencies in individual datasets

Request a detailed protocol

We wished to identify associations between driver gene mutations and gene sensitivity scores that were not confounded by tissue specific gene sensitivity effects (e.g. SOX10 sensitivity scores can be naively associated with BRAF mutational status because SOX10 is essential in melanoma cell lines and BRAF mutation is common in melanoma). Thus, we wished to model gene sensitivity after first accounting for tissue type. To this end, associations between individual driver genes and gene sensitivity scores were identified using an ANOVA model that incorporated both tissue type and mutational status as covariates, similar to the method previously developed for identifying pharmacogenomic interactions in cancer cell line panels (Cokelaer et al., 2018; Iorio et al., 2016). As recent work (Behan et al., 2019) has highlighted that some dependencies (e.g. WRN) can be associated with microsatellite instability rather than individual driver genes, we also incorporated microsatellite instability (Ghandi et al., 2019) as a covariate in our model. The model had the form ‘gene_X_sensitivity ~MSI_status + C(Tissue) + driver_gene_Y_status’ and was used to test the association between each recurrently mutated driver gene Y and all gene sensitivity scores X assayed in a given dataset. Driver genes were included in this analysis if they were functionally altered in at least five cell lines in the dataset being analysed. Correction for multiple hypothesis testing was performed using the Benjamini and Hochberg, 1995 false discovery rate.

Identifying genetic dependencies common to multiple datasets

Request a detailed protocol

When comparing a pair of datasets, we used one dataset as a discovery dataset and a second as a validation set, as outlined in Figure 1C. The discovery analysis was limited to the set of interactions that could be tested in both datasets, that is associations between the set of sensitivity scores for genes screened in both studies and the set of driver genes recurrently altered in both studies. An initial set of genetic interactions was identified in the discovery dataset at a specific FDR threshold and these associations were then tested in the validation set. We considered interactions to be reproduced in the validation dataset if: (1) the FDR was less than the threshold; (2) the uncorrected p-value was <0.05 and; (3) the sign of the association (sensitivity/resistance) was the same in both discovery and validation set. A FDR of 0.2 was used for all analysis presented in the main text but additional FDR thresholds (0.1, 0.3) were tested to ensure that all findings were robust to the exact choice of FDR (Figure 5—figure supplement 1).

Pathway enrichment of genetic dependencies

Request a detailed protocol

Pathway enrichment was assessed using gProfiler (Raudvere et al., 2019) with KEGG (Kanehisa et al., 2017) and Reactome (Jassal et al., 2020) as annotation databases and the selectively lethal genes as the background list.

Protein–protein interactions

Request a detailed protocol

Protein–protein interactions were obtained from STRING v10.5 (Szklarczyk et al., 2015) (RRID:SCR_005223), BIOGRID 3.5.170 (Chatr-Aryamontri et al., 2015) (RRID:SCR_007393), HIPPIE v2.0 (Alanis-Lobato et al., 2017)(RRID:SCR_014651) and BioPlex 3.0 (Huttlin et al., 2020) (RRID:SCR_016144). Results in the main text make use of medium confidence STRING interactions (STRING integrated score >0.4). However, to ensure robustness to the thresholds shown, all analyses were repeated for the full set of HIPPIE, BioGRID and BioPlex interactions (Figure 3—figure supplement 1). NetworkX version 2.2 (Hagberg et al., 2008)(RRID:SCR_016864) was used to create 100 randomised versions of the medium confidence STRING interaction network analysed in Figure 5—figure supplement 2.

siRNA experiments

Request a detailed protocol

HAP1 cell lines were obtained from Horizon Discovery: HAP1_WT (C631), HAP1_VPS4A_ KO (HZGHC004623c005), HAP1_VPS4B_KO (HZGHC006889c011), HAP1_DDX5_KO (HZGHC006136c012) and HAP1_DDX17_KO (HZGHC007221c009). Cells were cultured in IMDM (10–016-CV; Corning) supplemented with 10% FBS (10270–106; Thermo Fisher Scientific). ON-TARGETplus siRNA SMARTpools targeting VPS4A (L-013092-00-0005), VPS4B (L-013119-00-0005), DDX5 (L-003774-00-0005), DDX17 (L-013450-01-0005), PLK1 (L-003290-00-0005) and a non-targeting scramble control (D-001810-10-20) were obtained from Dharmacon. HAP1 cells were seeded to a density of 5000 cells per well of a 96-well plate and 5 nM siRNA was transfected with Lipofectamine 3000 (L3000015; Thermo Fisher Scientific) in Opti-MEM I Reduced Serum Medium (31985070; Thermo Fisher Scientific). Cell viability was measured 72 hr after siRNA transfection using CellTiter-Glo Luminescent Cell Viability Assay (G7570; Promega). The 96 well plates were read using a SpectraMax M3 Microplate Reader (Molecular devices). Viability effects for each siRNA targeting each gene X in each cell line y were normalised using the following formula:

NormalisedViability (siRNA_X_y)=1 - (siCTRL_y - siRNA_X_y) / (siCTRL_y - siPLK1_y) where siCTRL_y is the average of 3 measurements of non-targeting scramble control in cell line y and siPLK1_y is the average of 3 measurements of an siRNA smartpool targeting PLK1 in cell line y. Raw and normalised viability data are in Supplementary file 8.

Data availability

All data generated during this study are included in the manuscript and supporting files.

References

(2017) HIPPIE v2.0: enhancing meaningfulness and reliability of protein-protein interaction networks
Nucleic Acids Research 45:D408–D414.

https://doi.org/10.1093/nar/gkw985
- PubMed
- Google Scholar
(2015) The history and future of targeting cyclin-dependent kinases in Cancer therapy
Nature Reviews Drug Discovery 14:130–146.

https://doi.org/10.1038/nrd4504
- PubMed
- Google Scholar
1. Barretina J
2. Caponigro G
3. Stransky N
4. Venkatesan K
5. Margolin AA
6. Kim S
7. Wilson CJ
8. Lehár J
9. Kryukov GV
10. Sonkin D
11. Reddy A
12. Liu M
13. Murray L
14. Berger MF
15. Monahan JE
16. Morais P
17. Meltzer J
18. Korejwa A
19. Jané-Valbuena J
20. Mapa FA
21. Thibault J
22. Bric-Furlong E
23. Raman P
24. Shipway A
25. Engels IH
26. Cheng J
27. Yu GK
28. Yu J
29. Aspesi P
30. de Silva M
31. Jagtap K
32. Jones MD
33. Wang L
34. Hatton C
35. Palescandolo E
36. Gupta S
37. Mahan S
38. Sougnez C
39. Onofrio RC
40. Liefeld T
41. MacConaill L
42. Winckler W
43. Reich M
44. Li N
45. Mesirov JP
46. Gabriel SB
47. Getz G
48. Ardlie K
49. Chan V
50. Myer VE
51. Weber BL
52. Porter J
53. Warmuth M
54. Finan P
55. Harris JL
56. Meyerson M
57. Golub TR
58. Morrissey MP
59. Sellers WR
60. Schlegel R
61. Garraway LA
(2012) The Cancer cell line encyclopedia enables predictive modelling of anticancer drug sensitivity
Nature 483:603–607.

https://doi.org/10.1038/nature11003
- PubMed
- Google Scholar
1. Behan FM
2. Iorio F
3. Picco G
4. Gonçalves E
5. Beaver CM
6. Migliardi G
7. Santos R
8. Rao Y
9. Sassi F
10. Pinnelli M
11. Ansari R
12. Harper S
13. Jackson DA
14. McRae R
15. Pooley R
16. Wilkinson P
17. van der Meer D
18. Dow D
19. Buser-Doepner C
20. Bertotti A
21. Trusolino L
22. Stronach EA
23. Saez-Rodriguez J
24. Yusa K
25. Garnett MJ
(2019) Prioritization of Cancer therapeutic targets using CRISPR-Cas9 screens
Nature 568:511–516.

https://doi.org/10.1038/s41586-019-1103-9
- PubMed
- Google Scholar
1. Benjamini Y
2. Hochberg Y
(1995) Controlling the false discovery rate: a practical and powerful approach to multiple testing
Journal of the Royal Statistical Society: Series B 57:289–300.

https://doi.org/10.1111/j.2517-6161.1995.tb02031.x
- Google Scholar
1. Boehm JS
2. Golub TR
(2015) An ecosystem of Cancer cell line factories to support a Cancer dependency map
Nature Reviews Genetics 16:373–374.

https://doi.org/10.1038/nrg3967
- PubMed
- Google Scholar
1. Bridgett S
2. Campbell J
3. Lord CJ
4. Ryan CJ
(2017) CancerGD: a resource for identifying and interpreting genetic dependencies in Cancer
Cell Systems 5:82–86.

https://doi.org/10.1016/j.cels.2017.06.002
- PubMed
- Google Scholar
1. Brough R
2. Gulati A
3. Haider S
4. Kumar R
5. Campbell J
6. Knudsen E
7. Pettitt SJ
8. Ryan CJ
9. Lord CJ
(2018) Identification of highly penetrant Rb-related synthetic lethal interactions in triple negative breast Cancer
Oncogene 37:5701–5718.

https://doi.org/10.1038/s41388-018-0368-z
- PubMed
- Google Scholar
1. Campbell J
2. Ryan CJ
3. Brough R
4. Bajrami I
5. Pemberton HN
6. Chong IY
7. Costa-Cabral S
8. Frankum J
9. Gulati A
10. Holme H
11. Miller R
12. Postel-Vinay S
13. Rafiq R
14. Wei W
15. Williamson CT
16. Quigley DA
17. Tym J
18. Al-Lazikani B
19. Fenton T
20. Natrajan R
21. Strauss SJ
22. Ashworth A
23. Lord CJ
(2016) Large-Scale profiling of kinase dependencies in Cancer cell lines
Cell Reports 14:2490–2501.

https://doi.org/10.1016/j.celrep.2016.02.023
- PubMed
- Google Scholar
1. Cancer Cell Line Encyclopedia Consortium
2. Genomics of Drug Sensitivity in Cancer Consortium
(2015) Pharmacogenomic agreement between two Cancer cell line data sets
Nature 528:84–87.

https://doi.org/10.1038/nature15736
- PubMed
- Google Scholar
(2014) Synthetic lethal screens as a means to understand and treat MYC-driven cancers
Cold Spring Harbor Perspectives in Medicine 4:a014209.

https://doi.org/10.1101/cshperspect.a014209
- PubMed
- Google Scholar
1. Chapman PB
2. Hauschild A
3. Robert C
4. Haanen JB
5. Ascierto P
6. Larkin J
7. Dummer R
8. Garbe C
9. Testori A
10. Maio M
11. Hogg D
12. Lorigan P
13. Lebbe C
14. Jouary T
15. Schadendorf D
16. Ribas A
17. O'Day SJ
18. Sosman JA
19. Kirkwood JM
20. Eggermont AM
21. Dreno B
22. Nolop K
23. Li J
24. Nelson B
25. Hou J
26. Lee RJ
27. Flaherty KT
28. McArthur GA
29. BRIM-3 Study Group
(2011) Improved survival with vemurafenib in melanoma with BRAF V600E mutation
New England Journal of Medicine 364:2507–2516.

https://doi.org/10.1056/NEJMoa1103782
- PubMed
- Google Scholar
1. Chatr-Aryamontri A
2. Breitkreutz BJ
3. Oughtred R
4. Boucher L
5. Heinicke S
6. Chen D
7. Stark C
8. Breitkreutz A
9. Kolas N
10. O'Donnell L
11. Reguly T
12. Nixon J
13. Ramage L
14. Winter A
15. Sellam A
16. Chang C
17. Hirschman J
18. Theesfeld C
19. Rust J
20. Livstone MS
21. Dolinski K
22. Tyers M
(2015) The BioGRID interaction database: 2015 update
Nucleic Acids Research 43:D470–D478.

https://doi.org/10.1093/nar/gku1204
- PubMed
- Google Scholar
(2018) GDSCTools for mining pharmacogenomic interactions in Cancer
Bioinformatics 34:1226–1228.

https://doi.org/10.1093/bioinformatics/btx744
- PubMed
- Google Scholar
1. Costanzo M
2. Baryshnikova A
3. Bellay J
4. Kim Y
5. Spear ED
6. Sevier CS
7. Ding H
8. Koh JL
9. Toufighi K
10. Mostafavi S
11. Prinz J
12. St Onge RP
13. VanderSluis B
14. Makhnevych T
15. Vizeacoumar FJ
16. Alizadeh S
17. Bahr S
18. Brost RL
19. Chen Y
20. Cokol M
21. Deshpande R
22. Li Z
23. Lin ZY
24. Liang W
25. Marback M
26. Paw J
27. San Luis BJ
28. Shuteriqi E
29. Tong AH
30. van Dyk N
31. Wallace IM
32. Whitney JA
33. Weirauch MT
34. Zhong G
35. Zhu H
36. Houry WA
37. Brudno M
38. Ragibizadeh S
39. Papp B
40. Pál C
41. Roth FP
42. Giaever G
43. Nislow C
44. Troyanskaya OG
45. Bussey H
46. Bader GD
47. Gingras AC
48. Morris QD
49. Kim PM
50. Kaiser CA
51. Myers CL
52. Andrews BJ
53. Boone C
(2010) The genetic landscape of a cell
Science 327:425–431.

https://doi.org/10.1126/science.1180823
- PubMed
- Google Scholar
1. Dardenne E
2. Polay Espinoza M
3. Fattet L
4. Germann S
5. Lambert MP
6. Neil H
7. Zonta E
8. Mortada H
9. Gratadou L
10. Deygas M
11. Chakrama FZ
12. Samaan S
13. Desmet FO
14. Tranchevent LC
15. Dutertre M
16. Rimokh R
17. Bourgeois CF
18. Auboeuf D
(2014) RNA helicases DDX5 and DDX17 dynamically orchestrate transcription, miRNA, and splicing programs in cell differentiation
Cell Reports 7:1900–1913.

https://doi.org/10.1016/j.celrep.2014.05.010
- PubMed
- Google Scholar
1. Downward J
(2015) RAS synthetic lethal screens revisited: still seeking the elusive prize?
Clinical Cancer Research 21:1802–1809.

https://doi.org/10.1158/1078-0432.CCR-14-2180
- PubMed
- Google Scholar
1. Formstecher E
2. Ramos JW
3. Fauquet M
4. Calderwood DA
5. Hsieh JC
6. Canton B
7. Nguyen XT
8. Barnier JV
9. Camonis J
10. Ginsberg MH
11. Chneiweiss H
(2001) PEA-15 mediates cytoplasmic sequestration of ERK MAP kinase
Developmental Cell 1:239–250.

https://doi.org/10.1016/S1534-5807(01)00035-1
- PubMed
- Google Scholar
1. Fuller-Pace FV
(2013) DEAD box RNA helicase functions in Cancer
RNA Biology 10:121–132.

https://doi.org/10.4161/rna.23312
- PubMed
- Google Scholar
1. Ghandi M
2. Huang FW
3. Jané-Valbuena J
4. Kryukov GV
5. Lo CC
6. McDonald ER
7. Barretina J
8. Gelfand ET
9. Bielski CM
10. Li H
11. Hu K
12. Andreev-Drakhlin AY
13. Kim J
14. Hess JM
15. Haas BJ
16. Aguet F
17. Weir BA
18. Rothberg MV
19. Paolella BR
20. Lawrence MS
21. Akbani R
22. Lu Y
23. Tiv HL
24. Gokhale PC
25. de Weck A
26. Mansour AA
27. Oh C
28. Shih J
29. Hadi K
30. Rosen Y
31. Bistline J
32. Venkatesan K
33. Reddy A
34. Sonkin D
35. Liu M
36. Lehar J
37. Korn JM
38. Porter DA
39. Jones MD
40. Golji J
41. Caponigro G
42. Taylor JE
43. Dunning CM
44. Creech AL
45. Warren AC
46. McFarland JM
47. Zamanighomi M
48. Kauffmann A
49. Stransky N
50. Imielinski M
51. Maruvka YE
52. Cherniack AD
53. Tsherniak A
54. Vazquez F
55. Jaffe JD
56. Lane AA
57. Weinstock DM
58. Johannessen CM
59. Morrissey MP
60. Stegmeier F
61. Schlegel R
62. Hahn WC
63. Getz G
64. Mills GB
65. Boehm JS
66. Golub TR
67. Garraway LA
68. Sellers WR
(2019) Next-generation characterization of the Cancer cell line encyclopedia
Nature 569:503–508.

https://doi.org/10.1038/s41586-019-1186-3
- PubMed
- Google Scholar
1. Giacomelli AO
2. Yang X
3. Lintner RE
4. McFarland JM
5. Duby M
6. Kim J
7. Howard TP
8. Takeda DY
9. Ly SH
10. Kim E
11. Gannon HS
12. Hurhula B
13. Sharpe T
14. Goodale A
15. Fritchman B
16. Steelman S
17. Vazquez F
18. Tsherniak A
19. Aguirre AJ
20. Doench JG
21. Piccioni F
22. Roberts CWM
23. Meyerson M
24. Getz G
25. Johannessen CM
26. Root DE
27. Hahn WC
(2018) Mutational processes shape the landscape of TP53 mutations in human Cancer
Nature Genetics 50:1381–1387.

https://doi.org/10.1038/s41588-018-0204-y
- PubMed
- Google Scholar
Book
(2008)
Exploring Network Structure, Dynamics, and Function Using Networkx

Los Alamos, NM, United States: Los Alamos National Lab (LANL).
- Google Scholar
1. Haibe-Kains B
2. El-Hachem N
3. Birkbak NJ
4. Jin AC
5. Beck AH
6. Aerts HJ
7. Quackenbush J
(2013) Inconsistency in large pharmacogenomic studies
Nature 504:389–393.

https://doi.org/10.1038/nature12831
- PubMed
- Google Scholar
1. Han K
2. Jeng EE
3. Hess GT
4. Morgens DW
5. Li A
6. Bassik MC
(2017) Synergistic drug combinations for Cancer identified in a CRISPR screen for pairwise genetic interactions
Nature Biotechnology 35:463–474.

https://doi.org/10.1038/nbt.3834
- PubMed
- Google Scholar
1. Hart T
2. Tong AHY
3. Chan K
4. Van Leeuwen J
5. Seetharaman A
6. Aregger M
7. Chandrashekhar M
8. Hustedt N
9. Seth S
10. Noonan A
11. Habsid A
12. Sizova O
13. Nedyalkova L
14. Climie R
15. Tworzyanski L
16. Lawson K
17. Sartori MA
18. Alibeh S
19. Tieu D
20. Masud S
21. Mero P
22. Weiss A
23. Brown KR
24. Usaj M
25. Billmann M
26. Rahman M
27. Costanzo M
28. Myers CL
29. Andrews BJ
30. Boone C
31. Durocher D
32. Moffat J
(2017) Evaluation and design of Genome-Wide CRISPR/SpCas9 knockout screens
G3: Genes, Genomes, Genetics 7:2719–2727.

https://doi.org/10.1534/g3.117.041277
- Google Scholar
1. Hegele A
2. Kamburov A
3. Grossmann A
4. Sourlis C
5. Wowro S
6. Weimann M
7. Will CL
8. Pena V
9. Lührmann R
10. Stelzl U
(2012) Dynamic protein-protein interaction wiring of the human spliceosome
Molecular Cell 45:567–580.

https://doi.org/10.1016/j.molcel.2011.12.034
- PubMed
- Google Scholar
1. Helming KC
2. Wang X
3. Wilson BG
4. Vazquez F
5. Haswell JR
6. Manchester HE
7. Kim Y
8. Kryukov GV
9. Ghandi M
10. Aguirre AJ
11. Jagani Z
12. Wang Z
13. Garraway LA
14. Hahn WC
15. Roberts CW
(2014) ARID1B is a specific vulnerability in ARID1A-mutant cancers
Nature Medicine 20:251–254.

https://doi.org/10.1038/nm.3480
- PubMed
- Google Scholar
(2019) Context-dependent genetic interactions in Cancer
Current Opinion in Genetics & Development 54:73–82.

https://doi.org/10.1016/j.gde.2019.03.004
- PubMed
- Google Scholar
1. Herrero A
2. Pinto A
3. Colón-Bolea P
4. Casar B
5. Jones M
6. Agudo-Ibáñez L
7. Vidal R
8. Tenbaum SP
9. Nuciforo P
10. Valdizán EM
11. Horvath Z
12. Orfi L
13. Pineda-Lucena A
14. Bony E
15. Keri G
16. Rivas G
17. Pazos A
18. Gozalbes R
19. Palmer HG
20. Hurlstone A
21. Crespo P
(2015) Small molecule inhibition of ERK dimerization prevents tumorigenesis by RAS-ERK pathway oncogenes
Cancer Cell 28:170–182.

https://doi.org/10.1016/j.ccell.2015.07.001
- PubMed
- Google Scholar
1. Hill KS
2. Roberts ER
3. Wang X
4. Marin E
5. Park TD
6. Son S
7. Ren Y
8. Fang B
9. Yoder S
10. Kim S
11. Wan L
12. Sarnaik AA
13. Koomen JM
14. Messina JL
15. Teer JK
16. Kim Y
17. Wu J
18. Chalfant CE
19. Kim M
(2019) PTPN11 plays oncogenic roles and is a therapeutic target for BRAF Wild-Type Melanomas
Molecular Cancer Research 17:583–593.

https://doi.org/10.1158/1541-7786.MCR-18-0777
- PubMed
- Google Scholar
1. Hoffman GR
2. Rahal R
3. Buxton F
4. Xiang K
5. McAllister G
6. Frias E
7. Bagdasarian L
8. Huber J
9. Lindeman A
10. Chen D
11. Romero R
12. Ramadan N
13. Phadke T
14. Haas K
15. Jaskelioff M
16. Wilson BG
17. Meyer MJ
18. Saenz-Vash V
19. Zhai H
20. Myer VE
21. Porter JA
22. Keen N
23. McLaughlin ME
24. Mickanin C
25. Roberts CW
26. Stegmeier F
27. Jagani Z
(2014) Functional epigenetics approach identifies BRM/SMARCA2 as a critical synthetic lethal target in BRG1-deficient cancers
PNAS 111:3128–3133.

https://doi.org/10.1073/pnas.1316793111
- PubMed
- Google Scholar
1. Huttlin EL
2. Bruckner RJ
3. Paulo JA
4. Cannon JR
5. Ting L
6. Baltier K
7. Colby G
8. Gebreab F
9. Gygi MP
10. Parzen H
11. Szpyt J
12. Tam S
13. Zarraga G
14. Pontano-Vaites L
15. Swarup S
16. White AE
17. Schweppe DK
18. Rad R
19. Erickson BK
20. Obar RA
21. Guruharsha KG
22. Li K
23. Artavanis-Tsakonas S
24. Gygi SP
25. Harper JW
(2017) Architecture of the human interactome defines protein communities and disease networks
Nature 545:505–509.

https://doi.org/10.1038/nature22366
- PubMed
- Google Scholar
Preprint
1. Huttlin EL
2. Bruckner RJ
3. Navarrete-Perea J
4. Cannon JR
5. Baltier K
6. Gebreab F
7. Gygi MP
8. Thornock A
9. Zarraga G
10. Tam S
(2020) Dual proteome-scale networks reveal cell-specific remodeling of the human interactome
bioRxiv.

https://doi.org/10.1101/2020.01.19.905109
- Google Scholar
1. Iorio F
2. Knijnenburg TA
3. Vis DJ
4. Bignell GR
5. Menden MP
6. Schubert M
7. Aben N
8. Gonçalves E
9. Barthorpe S
10. Lightfoot H
11. Cokelaer T
12. Greninger P
13. van Dyk E
14. Chang H
15. de Silva H
16. Heyn H
17. Deng X
18. Egan RK
19. Liu Q
20. Mironenko T
21. Mitropoulos X
22. Richardson L
23. Wang J
24. Zhang T
25. Moran S
26. Sayols S
27. Soleimani M
28. Tamborero D
29. Lopez-Bigas N
30. Ross-Macdonald P
31. Esteller M
32. Gray NS
33. Haber DA
34. Stratton MR
35. Benes CH
36. Wessels LFA
37. Saez-Rodriguez J
38. McDermott U
39. Garnett MJ
(2016) A landscape of pharmacogenomic interactions in Cancer
Cell 166:740–754.

https://doi.org/10.1016/j.cell.2016.06.017
- PubMed
- Google Scholar
1. Iorio F
2. Behan FM
3. Gonçalves E
4. Bhosle SG
5. Chen E
6. Shepherd R
7. Beaver C
8. Ansari R
9. Pooley R
10. Wilkinson P
11. Harper S
12. Butler AP
13. Stronach EA
14. Saez-Rodriguez J
15. Yusa K
16. Garnett MJ
(2018) Unsupervised correction of gene-independent cell responses to CRISPR-Cas9 targeting
BMC Genomics 19:604.

https://doi.org/10.1186/s12864-018-4989-y
- PubMed
- Google Scholar
1. Jassal B
2. Matthews L
3. Viteri G
4. Gong C
5. Lorente P
6. Fabregat A
7. Sidiropoulos K
8. Cook J
9. Gillespie M
10. Haw R
11. Loney F
12. May B
13. Milacic M
14. Rothfels K
15. Sevilla C
16. Shamovsky V
17. Shorser S
18. Varusai T
19. Weiser J
20. Wu G
21. Stein L
22. Hermjakob H
23. D'Eustachio P
(2020) The reactome pathway knowledgebase
Nucleic Acids Research 48:D498–D503.

https://doi.org/10.1093/nar/gkz1031
- PubMed
- Google Scholar
1. Jerby-Arnon L
2. Pfetzer N
3. Waldman YY
4. McGarry L
5. James D
6. Shanks E
7. Seashore-Ludlow B
8. Weinstock A
9. Geiger T
10. Clemons PA
11. Gottlieb E
12. Ruppin E
(2014) Predicting cancer-specific vulnerability via data-driven detection of synthetic lethality
Cell 158:1199–1209.

https://doi.org/10.1016/j.cell.2014.07.027
- PubMed
- Google Scholar
1. Ji P
2. Jiang H
3. Rekhtman K
4. Bloom J
5. Ichetovkin M
6. Pagano M
7. Zhu L
(2004) An Rb-Skp2-p27 pathway mediates acute cell cycle inhibition by rb and is retained in a partial-penetrance rb mutant
Molecular Cell 16:47–58.

https://doi.org/10.1016/j.molcel.2004.09.029
- PubMed
- Google Scholar
1. Kaelin WG
(2012) Use and abuse of RNAi to study mammalian gene function
Science 337:421–422.

https://doi.org/10.1126/science.1225787
- Google Scholar
(2017) KEGG: new perspectives on genomes, pathways, diseases and drugs
Nucleic Acids Research 45:D353–D361.

https://doi.org/10.1093/nar/gkw1092
- PubMed
- Google Scholar
1. Kelley R
2. Ideker T
(2005) Systematic interpretation of genetic interactions using protein networks
Nature Biotechnology 23:561–566.

https://doi.org/10.1038/nbt1096
- PubMed
- Google Scholar
1. Kryukov GV
2. Wilson FH
3. Ruth JR
4. Paulk J
5. Tsherniak A
6. Marlow SE
7. Vazquez F
8. Weir BA
9. Fitzgerald ME
10. Tanaka M
11. Bielski CM
12. Scott JM
13. Dennis C
14. Cowley GS
15. Boehm JS
16. Root DE
17. Golub TR
18. Clish CB
19. Bradner JE
20. Hahn WC
21. Garraway LA
(2016) MTAP deletion confers enhanced dependency on the PRMT5 arginine methyltransferase in Cancer cells
Science 351:1214–1218.

https://doi.org/10.1126/science.aad5214
- PubMed
- Google Scholar
1. Lord CJ
2. Ashworth A
(2017) PARP inhibitors: synthetic lethality in the clinic
Science 355:1152–1158.

https://doi.org/10.1126/science.aam7344
- PubMed
- Google Scholar
1. Marcotte R
2. Sayad A
3. Brown KR
4. Sanchez-Garcia F
5. Reimand J
6. Haider M
7. Virtanen C
8. Bradner JE
9. Bader GD
10. Mills GB
11. Pe'er D
12. Moffat J
13. Neel BG
(2016) Functional genomic landscape of human breast Cancer drivers, vulnerabilities, and resistance
Cell 164:293–309.

https://doi.org/10.1016/j.cell.2015.11.062
- PubMed
- Google Scholar
1. Marjon K
2. Cameron MJ
3. Quang P
4. Clasquin MF
5. Mandley E
6. Kunii K
7. McVay M
8. Choe S
9. Kernytsky A
10. Gross S
11. Konteatis Z
12. Murtie J
13. Blake ML
14. Travins J
15. Dorsch M
16. Biller SA
17. Marks KM
(2016) MTAP deletions in Cancer create vulnerability to targeting of the MAT2A/PRMT5/RIOK1 Axis
Cell Reports 15:574–587.

https://doi.org/10.1016/j.celrep.2016.03.043
- PubMed
- Google Scholar
1. Martin TD
2. Cook DR
3. Choi MY
4. Li MZ
5. Haigis KM
6. Elledge SJ
(2017) A role for mitochondrial translation in promotion of viability in K-Ras mutant cells
Cell Reports 20:427–438.

https://doi.org/10.1016/j.celrep.2017.06.061
- PubMed
- Google Scholar
1. Mavrakis KJ
2. McDonald ER
3. Schlabach MR
4. Billy E
5. Hoffman GR
6. deWeck A
7. Ruddy DA
8. Venkatesan K
9. Yu J
10. McAllister G
11. Stump M
12. deBeaumont R
13. Ho S
14. Yue Y
15. Liu Y
16. Yan-Neale Y
17. Yang G
18. Lin F
19. Yin H
20. Gao H
21. Kipp DR
22. Zhao S
23. McNamara JT
24. Sprague ER
25. Zheng B
26. Lin Y
27. Cho YS
28. Gu J
29. Crawford K
30. Ciccone D
31. Vitari AC
32. Lai A
33. Capka V
34. Hurov K
35. Porter JA
36. Tallarico J
37. Mickanin C
38. Lees E
39. Pagliarini R
40. Keen N
41. Schmelzle T
42. Hofmann F
43. Stegmeier F
44. Sellers WR
(2016) Disordered methionine metabolism in MTAP/CDKN2A-deleted cancers leads to dependence on PRMT5
Science 351:1208–1213.

https://doi.org/10.1126/science.aad5944
- PubMed
- Google Scholar
1. McDonald ER
2. de Weck A
3. Schlabach MR
4. Billy E
5. Mavrakis KJ
6. Hoffman GR
7. Belur D
8. Castelletti D
9. Frias E
10. Gampa K
11. Golji J
12. Kao I
13. Li L
14. Megel P
15. Perkins TA
16. Ramadan N
17. Ruddy DA
18. Silver SJ
19. Sovath S
20. Stump M
21. Weber O
22. Widmer R
23. Yu J
24. Yu K
25. Yue Y
26. Abramowski D
27. Ackley E
28. Barrett R
29. Berger J
30. Bernard JL
31. Billig R
32. Brachmann SM
33. Buxton F
34. Caothien R
35. Caushi JX
36. Chung FS
37. Cortés-Cros M
38. deBeaumont RS
39. Delaunay C
40. Desplat A
41. Duong W
42. Dwoske DA
43. Eldridge RS
44. Farsidjani A
45. Feng F
46. Feng J
47. Flemming D
48. Forrester W
49. Galli GG
50. Gao Z
51. Gauter F
52. Gibaja V
53. Haas K
54. Hattenberger M
55. Hood T
56. Hurov KE
57. Jagani Z
58. Jenal M
59. Johnson JA
60. Jones MD
61. Kapoor A
62. Korn J
63. Liu J
64. Liu Q
65. Liu S
66. Liu Y
67. Loo AT
68. Macchi KJ
69. Martin T
70. McAllister G
71. Meyer A
72. Mollé S
73. Pagliarini RA
74. Phadke T
75. Repko B
76. Schouwey T
77. Shanahan F
78. Shen Q
79. Stamm C
80. Stephan C
81. Stucke VM
82. Tiedt R
83. Varadarajan M
84. Venkatesan K
85. Vitari AC
86. Wallroth M
87. Weiler J
88. Zhang J
89. Mickanin C
90. Myer VE
91. Porter JA
92. Lai A
93. Bitter H
94. Lees E
95. Keen N
96. Kauffmann A
97. Stegmeier F
98. Hofmann F
99. Schmelzle T
100. Sellers WR
(2017) Project DRIVE: a compendium of Cancer dependencies and synthetic lethal relationships uncovered by Large-Scale, deep RNAi screening
Cell 170:577–592.

https://doi.org/10.1016/j.cell.2017.07.005
- PubMed
- Google Scholar
Conference
1. McKinney W
(2011)
Pandas: a foundational Python library for data analysis and statistics

Python for High Performance and Scientific Computing.
- Google Scholar
1. Meyers RM
2. Bryan JG
3. McFarland JM
4. Weir BA
5. Sizemore AE
6. Xu H
7. Dharia NV
8. Montgomery PG
9. Cowley GS
10. Pantel S
11. Goodale A
12. Lee Y
13. Ali LD
14. Jiang G
15. Lubonja R
16. Harrington WF
17. Strickland M
18. Wu T
19. Hawes DC
20. Zhivich VA
21. Wyatt MR
22. Kalani Z
23. Chang JJ
24. Okamoto M
25. Stegmaier K
26. Golub TR
27. Boehm JS
28. Vazquez F
29. Root DE
30. Hahn WC
31. Tsherniak A
(2017) Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in Cancer cells
Nature Genetics 49:1779–1784.

https://doi.org/10.1038/ng.3984
- PubMed
- Google Scholar
1. Michaut M
2. Baryshnikova A
3. Costanzo M
4. Myers CL
5. Andrews BJ
6. Boone C
7. Bader GD
(2011) Protein complexes are central in the yeast genetic landscape
PLOS Computational Biology 7:e1001092.

https://doi.org/10.1371/journal.pcbi.1001092
- PubMed
- Google Scholar
1. Muller FL
2. Colla S
3. Aquilanti E
4. Manzo VE
5. Genovese G
6. Lee J
7. Eisenson D
8. Narurkar R
9. Deng P
10. Nezi L
11. Lee MA
12. Hu B
13. Hu J
14. Sahin E
15. Ong D
16. Fletcher-Sananikone E
17. Ho D
18. Kwong L
19. Brennan C
20. Wang YA
21. Chin L
22. DePinho RA
(2012) Passenger deletions generate therapeutic vulnerabilities in Cancer
Nature 488:337–342.

https://doi.org/10.1038/nature11331
- PubMed
- Google Scholar
(2015) Collateral lethality: a new therapeutic strategy in oncology
Trends in Cancer 1:161–173.

https://doi.org/10.1016/j.trecan.2015.10.002
- PubMed
- Google Scholar
1. Niepel M
2. Hafner M
3. Mills CE
4. Subramanian K
5. Williams EH
6. Chung M
7. Gaudio B
8. Barrette AM
9. Stern AD
10. Hu B
11. Korkola JE
12. LINCS Consortium
13. Gray JW
14. Birtwistle MR
15. Heiser LM
16. Sorger PK
(2019) A Multi-center study on the reproducibility of Drug-Response assays in mammalian cell lines
Cell Systems 9:35–48.

https://doi.org/10.1016/j.cels.2019.06.005
- PubMed
- Google Scholar
1. O'Leary B
2. Cutts RJ
3. Liu Y
4. Hrebien S
5. Huang X
6. Fenwick K
7. André F
8. Loibl S
9. Loi S
10. Garcia-Murillas I
11. Cristofanilli M
12. Huang Bartlett C
13. Turner NC
(2018) The genetic landscape and clonal evolution of breast Cancer resistance to palbociclib plus fulvestrant in the PALOMA-3 trial
Cancer Discovery 8:1390–1403.

https://doi.org/10.1158/2159-8290.CD-18-0264
- PubMed
- Google Scholar
1. Oike T
2. Ogiwara H
3. Tominaga Y
4. Ito K
5. Ando O
6. Tsuta K
7. Mizukami T
8. Shimada Y
9. Isomura H
10. Komachi M
11. Furuta K
12. Watanabe S
13. Nakano T
14. Yokota J
15. Kohno T
(2013) A synthetic lethality-based strategy to treat cancers harboring a genetic deficiency in the chromatin remodeling factor BRG1
Cancer Research 73:5508–5518.

https://doi.org/10.1158/0008-5472.CAN-12-4593
- PubMed
- Google Scholar
1. Raudvere U
2. Kolberg L
3. Kuzmin I
4. Arak T
5. Adler P
6. Peterson H
7. Vilo J
(2019) G:profiler: a web server for functional enrichment analysis and conversions of gene lists (2019 update)
Nucleic Acids Research 47:W191–W198.

https://doi.org/10.1093/nar/gkz369
- PubMed
- Google Scholar
1. Roguev A
2. Bandyopadhyay S
3. Zofall M
4. Zhang K
5. Fischer T
6. Collins SR
7. Qu H
8. Shales M
9. Park HO
10. Hayles J
11. Hoe KL
12. Kim DU
13. Ideker T
14. Grewal SI
15. Weissman JS
16. Krogan NJ
(2008) Conservation and rewiring of functional modules revealed by an epistasis map in fission yeast
Science 322:405–410.

https://doi.org/10.1126/science.1162609
- PubMed
- Google Scholar
1. Roguev A
2. Talbot D
3. Negri GL
4. Shales M
5. Cagney G
6. Bandyopadhyay S
7. Panning B
8. Krogan NJ
(2013) Quantitative genetic-interaction mapping in mammalian cells
Nature Methods 10:432–437.

https://doi.org/10.1038/nmeth.2398
- PubMed
- Google Scholar
1. Rolland T
2. Taşan M
3. Charloteaux B
4. Pevzner SJ
5. Zhong Q
6. Sahni N
7. Yi S
8. Lemmens I
9. Fontanillo C
10. Mosca R
11. Kamburov A
12. Ghiassian SD
13. Yang X
14. Ghamsari L
15. Balcha D
16. Begg BE
17. Braun P
18. Brehme M
19. Broly MP
20. Carvunis AR
21. Convery-Zupan D
22. Corominas R
23. Coulombe-Huntington J
24. Dann E
25. Dreze M
26. Dricot A
27. Fan C
28. Franzosa E
29. Gebreab F
30. Gutierrez BJ
31. Hardy MF
32. Jin M
33. Kang S
34. Kiros R
35. Lin GN
36. Luck K
37. MacWilliams A
38. Menche J
39. Murray RR
40. Palagi A
41. Poulin MM
42. Rambout X
43. Rasla J
44. Reichert P
45. Romero V
46. Ruyssinck E
47. Sahalie JM
48. Scholz A
49. Shah AA
50. Sharma A
51. Shen Y
52. Spirohn K
53. Tam S
54. Tejeda AO
55. Trigg SA
56. Twizere JC
57. Vega K
58. Walsh J
59. Cusick ME
60. Xia Y
61. Barabási AL
62. Iakoucheva LM
63. Aloy P
64. De Las Rivas J
65. Tavernier J
66. Calderwood MA
67. Hill DE
68. Hao T
69. Roth FP
70. Vidal M
(2014) A proteome-scale map of the human interactome network
Cell 159:1212–1226.

https://doi.org/10.1016/j.cell.2014.10.050
- PubMed
- Google Scholar
1. Ryan CJ
2. Roguev A
3. Patrick K
4. Xu J
5. Jahari H
6. Tong Z
7. Beltrao P
8. Shales M
9. Qu H
10. Collins SR
11. Kliegman JI
12. Jiang L
13. Kuo D
14. Tosti E
15. Kim HS
16. Edelmann W
17. Keogh MC
18. Greene D
19. Tang C
20. Cunningham P
21. Shokat KM
22. Cagney G
23. Svensson JP
24. Guthrie C
25. Espenshade PJ
26. Ideker T
27. Krogan NJ
(2012) Hierarchical modularity and the evolution of genetic interactomes across species
Molecular Cell 46:691–704.

https://doi.org/10.1016/j.molcel.2012.05.028
- PubMed
- Google Scholar
(2018) Synthetic lethality and Cancer - Penetrance as the major barrier
Trends in Cancer 4:671–683.

https://doi.org/10.1016/j.trecan.2018.08.003
- PubMed
- Google Scholar
Conference
1. Seabold S
2. Perktold J
(2010)
Statsmodels: econometric and statistical modeling with Python

Proceedings of the 9th Python in Science Conference (SciPy Society Austin).
- Google Scholar
1. Shao DD
2. Tsherniak A
3. Gopal S
4. Weir BA
5. Tamayo P
6. Stransky N
7. Schumacher SE
8. Zack TI
9. Beroukhim R
10. Garraway LA
11. Margolin AA
12. Root DE
13. Hahn WC
14. Mesirov JP
(2013) ATARiS: computational quantification of gene suppression phenotypes from multisample RNAi screens
Genome Research 23:665–678.

https://doi.org/10.1101/gr.143586.112
- PubMed
- Google Scholar
1. Shen JP
2. Zhao D
3. Sasik R
4. Luebeck J
5. Birmingham A
6. Bojorquez-Gomez A
7. Licon K
8. Klepper K
9. Pekin D
10. Beckett AN
11. Sanchez KS
12. Thomas A
13. Kuo CC
14. Du D
15. Roguev A
16. Lewis NE
17. Chang AN
18. Kreisberg JF
19. Krogan N
20. Qi L
21. Ideker T
22. Mali P
(2017) Combinatorial CRISPR-Cas9 screens for de novo mapping of genetic interactions
Nature Methods 14:573–576.

https://doi.org/10.1038/nmeth.4225
- PubMed
- Google Scholar
1. Srivas R
2. Shen JP
3. Yang CC
4. Sun SM
5. Li J
6. Gross AM
7. Jensen J
8. Licon K
9. Bojorquez-Gomez A
10. Klepper K
11. Huang J
12. Pekin D
13. Xu JL
14. Yeerna H
15. Sivaganesh V
16. Kollenstart L
17. van Attikum H
18. Aza-Blanc P
19. Sobol RW
20. Ideker T
(2016) A network of conserved synthetic lethal interactions for exploration of precision Cancer therapy
Molecular Cell 63:514–525.

https://doi.org/10.1016/j.molcel.2016.06.022
- PubMed
- Google Scholar
1. Steckel M
2. Molina-Arcas M
3. Weigelt B
4. Marani M
5. Warne PH
6. Kuznetsov H
7. Kelly G
8. Saunders B
9. Howell M
10. Downward J
11. Hancock DC
(2012) Determination of synthetic lethal interactions in KRAS oncogene-dependent Cancer cells reveals novel therapeutic targeting strategies
Cell Research 22:1227–1245.

https://doi.org/10.1038/cr.2012.82
- PubMed
- Google Scholar
1. Szklarczyk D
2. Franceschini A
3. Wyder S
4. Forslund K
5. Heller D
6. Huerta-Cepas J
7. Simonovic M
8. Roth A
9. Santos A
10. Tsafou KP
11. Kuhn M
12. Bork P
13. Jensen LJ
14. von Mering C
(2015) STRING v10: protein-protein interaction networks, integrated over the tree of life
Nucleic Acids Research 43:D447–D452.

https://doi.org/10.1093/nar/gku1003
- PubMed
- Google Scholar
1. Tadesse S
2. Caldon EC
3. Tilley W
4. Wang S
(2019) Cyclin-Dependent kinase 2 inhibitors in Cancer therapy: an update
Journal of Medicinal Chemistry 62:4233–4251.

https://doi.org/10.1021/acs.jmedchem.8b01469
- PubMed
- Google Scholar
1. Tong AH
2. Evangelista M
3. Parsons AB
4. Xu H
5. Bader GD
6. Pagé N
7. Robinson M
8. Raghibizadeh S
9. Hogue CW
10. Bussey H
11. Andrews B
12. Tyers M
13. Boone C
(2001) Systematic genetic analysis with ordered arrays of yeast deletion mutants
Science 294:2364–2368.

https://doi.org/10.1126/science.1065810
- PubMed
- Google Scholar
1. Tsherniak A
2. Vazquez F
3. Montgomery PG
4. Weir BA
5. Kryukov G
6. Cowley GS
7. Gill S
8. Harrington WF
9. Pantel S
10. Krill-Burger JM
11. Meyers RM
12. Ali L
13. Goodale A
14. Lee Y
15. Jiang G
16. Hsiao J
17. Gerath WFJ
18. Howell S
19. Merkel E
20. Ghandi M
21. Garraway LA
22. Root DE
23. Golub TR
24. Boehm JS
25. Hahn WC
(2017) Defining a Cancer dependency map
Cell 170:564–576.

https://doi.org/10.1016/j.cell.2017.06.010
- PubMed
- Google Scholar
1. van der Meer D
2. Barthorpe S
3. Yang W
4. Lightfoot H
5. Hall C
6. Gilbert J
7. Francies HE
8. Garnett MJ
(2019) Cell model Passports-a hub for clinical, genetic and functional datasets of preclinical Cancer models
Nucleic Acids Research 47:D923–D929.

https://doi.org/10.1093/nar/gky872
- PubMed
- Google Scholar
(2015) The orally active and bioavailable ATR kinase inhibitor AZD6738 potentiates the anti-tumor effects of cisplatin to resolve ATM-deficient non-small cell lung Cancer in vivo
Oncotarget 6:44289–44305.

https://doi.org/10.18632/oncotarget.6247
- PubMed
- Google Scholar
(2013) Cancer genome landscapes
Science 339:1546–1558.

https://doi.org/10.1126/science.1235122
- PubMed
- Google Scholar
1. Wang C
2. Wang G
3. Feng X
4. Shepherd P
5. Zhang J
6. Tang M
7. Chen Z
8. Srivastava M
9. McLaughlin ME
10. Navone NM
11. Hart GT
12. Chen J
(2019) Genome-wide CRISPR screens reveal synthetic lethality of RNASEH2 deficiency and ATR inhibition
Oncogene 38:2451–2463.

https://doi.org/10.1038/s41388-018-0606-4
- PubMed
- Google Scholar
1. Wee S
2. Wiederschain D
3. Maira SM
4. Loo A
5. Miller C
6. deBeaumont R
7. Stegmeier F
8. Yao YM
9. Lengauer C
(2008) PTEN-deficient cancers depend on PIK3CB
PNAS 105:13057–13062.

https://doi.org/10.1073/pnas.0802655105
- PubMed
- Google Scholar
1. Wilson BG
2. Roberts CW
(2011) SWI/SNF nucleosome remodellers and Cancer
Nature Reviews Cancer 11:481–492.

https://doi.org/10.1038/nrc3068
- PubMed
- Google Scholar
1. Xu XL
2. Singh HP
3. Wang L
4. Qi DL
5. Poulos BK
6. Abramson DH
7. Jhanwar SC
8. Cobrinik D
(2014) Rb suppresses human cone-precursor-derived retinoblastoma tumours
Nature 514:385–388.

https://doi.org/10.1038/nature13813
- PubMed
- Google Scholar

Article and author information

Author details

Christopher J Lord

Breast Cancer Now Toby Robins Research Centre and Cancer Research UK Gene Function Laboratory, Institute of Cancer Research, London, United Kingdom

Contribution
Conceptualization, Funding acquisition, Writing - original draft, Writing - review and editing

Competing interests
No competing interests declared
Niall Quinn

School of Computer Science and Systems Biology Ireland, University College Dublin, Dublin, Ireland

Contribution
Validation, Investigation

Competing interests
No competing interests declared
Colm J Ryan

School of Computer Science and Systems Biology Ireland, University College Dublin, Dublin, Ireland

Contribution
Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Visualization, Methodology, Writing - original draft, Writing - review and editing

For correspondence
colm.ryan@ucd.ie

Competing interests
No competing interests declared

"This ORCID iD identifies the author of this article:" 0000-0003-2750-9854

Funding

Irish Research Council (Laureate Awards 2017/2018)

Colm J Ryan

Cancer Research UK (CRUK/A14276)

Christopher J Lord

Breast Cancer Now (CTR-Q4-Y2)

Christopher J Lord

Wellcome Trust (SFI-HRB-Wellcome Trust Biomedical Research Partnership 103049/Z/13/Z)

Colm J Ryan

Science Foundation Ireland (SFI-HRB-Wellcome Trust Biomedical Research Partnership 103049/Z/13/Z)

Colm J Ryan

Health Research Board (SFI-HRB-Wellcome Trust Biomedical Research Partnership 103049/Z/13/Z)

Colm J Ryan

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

CJL’s work on this study was funded by Breast Cancer Now as part of Programme Funding to the Breast Cancer Now Toby Robins Research Centre (CTR-Q4-Y2) and via a Cancer Research UK Programme Grant (CRUK/A14276). CJR’s work on this study was funded by the HRB/SFI/Wellcome Trust partnership (grant number 103049/Z/13/Z) and by the Irish Research Council Laureate Awards 2017/2018. We thank members of the Ryan and Lord labs for helpful comments and discussions.

Copyright

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.