Mathys et al., 2019 undertook the first single-nucleus RNA-seq (snRNA-seq) study of Alzheimer’s disease (AD). The authors profiled the transcriptomes of approximately 80,000 cells from the prefrontal cortex, collected from 48 individuals – 24 of which presented with varying degrees of AD pathology. (Mathys et al., 2019) data processing and quality control (QC) strategy for their snRNA-seq data was state of the art at this time. Furthermore, the authors took extra measures in an attempt to ensure the reliability of their results. Here, we reanalyse this data as not a criticism of the study, but as an endeavour to raise awareness and provide recommendations for rigorous analysis of single-cell and single-nucleus RNA-seq data (sc/snRNA-seq) for future studies. Most importantly, we aim to ensure that the AD research field does not focus on spuriously identified genes.

Our questions of Mathys et al., 2019 focus around their data processing and their differential expression (DE) analysis (Figure 1). Firstly, in relation to their processing approach, the authors discussed the high percentages of mitochondrial reads and low number of reads per cell present in their data. This is indicative of low cell quality (Ilicic et al., 2016); however, we believe the authors’ QC approach may not capture all of these low-quality cells. Moreover, the authors did not integrate the cells from different individuals to account for batch effects. As the field has matured since the authors’ work was published, dataset integration has become a common step in sc-RNA-seq protocols and is recommended by some to remove confounding sources of variation (Heumos et al., 2023; Amezquita et al., 2020; Tran et al., 2020). To gain advantage of these recent approaches, we used scFlow (Khozoie et al., 2021) to reprocess the authors’ data. This pipeline included the removal of empty droplets, nuclei with low read counts and doublets, followed by embedding and integration of cells from separate samples and cell typing. scFlow combines best-practice approaches for processing sc/snRNA-seq datasets; see ‘Materials and methods’ for a detailed explanation of these steps. Reprocessing resulted in 50,831 cells passing QC, approximately 20,000 less than the authors’ postprocessing set with differing cell-type proportions (Figures 2 and 3).

Figure 1

Download asset Open asset

Figure 2

Download asset Open asset

Figure 3

Download asset Open asset

With regards to data quality, it is worth noting that over 99% of nuclei had less than 200 genes expressed (Table 1). While this QC step was not unique to our reprocessing, the authors made the same exclusion in their analysis (Mathys et al., 2019), it highlights the relatively low quality of the data which may be attributable to the early stage of snRNA-seq technology of the time. For example, Brase et al.’s recent study of snRNA-seq of autosomal-dominant AD (Brase et al., 2023) used a more stringent cut-off for the minimum number of expressed genes and still kept 27% (122 times more) of the assayed cells after all QC steps. Moreover, the authors discussed the high percentages of mitochondrial reads in their data. The differences in approaches to filtering based on the proportion mitochondrial reads accounts for the notable discrepancy in the number of nuclei after QC between our approach and the authors’. Our approach used a 10% cut-off for the proportion of mitochondrial reads in a nuclei, as set out in Amezquita et al.’s best-practice guidelines (Amezquita et al., 2020), which is less stringent than Seurat’s guidelines (5%) (Hao et al., 2021) or that from Heumos et al., 2023 (8% from a median absolute deviations [MAD]-based cut-off selection). Conversely, the authors filtered out high mitochondrial read nuclei based on clusters from their t-SNE projection of the data (Mathys et al., 2019). Even at our lenient cut-off, over 16,000 nuclei that were removed in our QC pipeline were kept by the authors’ Figure 2, explaining the discrepancy in the number of nuclei after QC. Based on Figure 2, it is clear that the authors’ approach was ineffective at removing nuclei with high proportions of mitochondrial reads which is indicative of cell death (Heumos et al., 2023; Ilicic et al., 2016) – both excitatory and inhibitory nuclei with higher than 75% reads from the mitochondria were kept in the final processed dataset by the authors. We have made the data from our alternative processing approach publicly available (through Synapse: https://doi.org/10.7303/syn51758062.1) so that researchers can utilise this resource free of low-quality nuclei.

Our second question of Mathys et al., 2019 is their DE approach. The authors conducted a DE analysis between the controls and the patients with AD pathology, concentrating on six neuronal and glial cell types; excitatory neurons, inhibitory neurons, astrocytes, microglia, oligodendrocytes, and oligodendrocyte precursor cells, derived from the Allen Brain Atlas (Tasic et al., 2018). They performed downstream analysis on their identified differentially expressed genes (DEGs) and investigated some of the most compelling genes in more detail. Therefore, all findings put forward by their paper were based upon the validity of their DE approach. However, for this approach, the authors conducted a two-part, cell- and patient-level analysis. The cell-level analysis took each cell as an independent replicate, and the results of which were compared for consistency in directionality and rank of their DEGs against their patient-level analysis, a Poisson mixed model. The authors identified 1031 DEGs using this combinatorial approach – DEGs requiring a false discovery rate (FDR) < 0.01 in the cell-level and an FDR < 0.05 in the patient-level analysis. It is important to note that this cell-level DE approach, also known as pseudoreplication, overestimates the confidence in DEGs due to the statistical dependence between cells from the same patient not being considered (Murphy and Skene, 2022; Squair et al., 2021; Zimmerman et al., 2021; Lazic, 2010). When we inspect all DEGs identified at an FDR of 0.05 from the authors’ cell-level analysis, this number increases to 14,274. Pseudobulk DE analysis has recently been proven to give optimal performance compared to both mixed models and pseudoreplication approaches (Murphy and Skene, 2022; Squair et al., 2021; Crowell et al., 2020; Soneson and Robinson, 2018). It aggregates counts to individuals, thus accounting for the dependence between an individual’s cells.

Here, to compare the effect of the different DE approaches in isolation, we apply a pseudobulk DE approach (Chen et al., 2016) to the authors’ original processed data. We found 26 unique DEGs when considering the six cell types used by the authors (Table 2). This was 549 times fewer DEGs than that reported originally at an FDR of 0.05. When we compare these DEGs, we can see that the absolute log₂ fold change (LFC) of our DEGs is 15 times larger than the authors’; median LFC of 2.34 and 0.16, despite the authors’ DEGs having an FDR score 8000 times smaller; median FDR of 2.89 × 10^–7and 0.002 (Figure 1a and b). Although we examined a high correlation in the genes’ fold change values across our pseudobulk analysis and the authors’ pseudoreplication analysis (Pearson R of 0.87 for an adjusted p-value of 0.05, Table 3), the p-values and resulting DEGs vary considerably. The correspondence in fold change values is expected given the approaches are applied to the same dataset, whereas the probabilities, which pertain to the likelihood that a gene’s expressional changes is related to the case/control differences in AD, importantly do not align. We can show that this stark contrast is just an artefact of the authors taking cells as independent replicates and thus artificially inflating confidence by considering the Pearson correlation between the number of DEGs found and the cell counts (Figure 1c–e). There is a near perfect, positive correlation between DEG and cell counts for the authors’ pseudoreplication analysis (Figure 1c) and for the 1031 genes from the authors’ combinatorial approach (Figure 1d) which is not present in our pseudobulk reanalysis (Figure 1e).

A further point which questions the authors’ DE approach is that they identified the vast majority of DEGs in the more abundant, neuronal cell types (Mathys et al., 2019). However, an increase in the number of cells is not the same as an increase in sample size since these cells are not independent from one another – they come from the same sample. Therefore, an increase in the number of cells should not necessarily result in an increase in the number of DEGs, whereas an increase in the number of samples would. This point is the major issue with pseudoreplication approaches which overestimate confidence when performing DE due to the statistical dependence between cells from the same patient not being considered (Squair et al., 2021; Lazic, 2010). In our opinion, it makes more sense to identify the majority of large effect size DEGs in microglia which recent work has established is the primary cell type by which the genetic risk for AD acts (Skene and Grant, 2016; McQuade and Blurton-Jones, 2019). This is what we found with our pseudobulk DE approach – 96% of all DEGs were in microglia (Table 2), whereas only 3% of the authors’ DEGs were in microglia.

Although it has been proven that pseudoreplication approaches result in false positives by artificially inflating the confidence from non-independent samples, we wanted to investigate the effect of the approach on the authors’ dataset. We ran the same cell-level analysis approach – a Wilcoxon rank-sum test and FDR multiple-testing correction – 100 times whilst randomly permuting the patient identifiers (Figure 1f). We would expect to find minimal DEGs with this approach given the random mixing of case and control patients. However, this pseudoreplication approach consistently found high numbers of DEGs, and we observe the same correlation between the number of cells and the number of DEGs as with the authors’ results. We did not observe the same pattern when running the same analysis with pseudobulk DE (Figure 1g). As a result, we conclude that integrating this pseudoreplication approach with a mixed model like the authors proposed just artificially inflates the test confidence for a random sample of the genes resulting in more false discoveries in cell types with bigger counts.

Up to this point, to compare the effect of the DE approaches in isolation, we analysed the same processed data from the authors as opposed to our reprocessed data. We also performed pseudobulk DE on our reprocessed data and found 16 unique DEGs (Table 4). It is worth noting that the fold change correlation between our two DE analyses (reprocessed data vs authors’ processed data) on the identified DEGs is only moderate (Pearson R of 0.57) and is lower than that of the correlation between pseudoreplication and pseudobulk on the same dataset (Table 3). This highlights the effect that the low quality high mitochondrial read cells have on DE analysis.

In conclusion, the authors’ analysis has been highly influential in the field with numerous studies undertaken based on their results, something we show has uncertain foundations. However, we would like to highlight that the use of pseudoreplication in neuroscience research is not isolated to the authors’ work; others have used this approach (Fernandes et al., 2020; Lui et al., 2021; Wakhloo et al., 2020), and their results should be similarly scrutinised. Here, we provide our processed count matrix with metadata and also the DEGs identified using an independently validated, DE approach so that other researchers can use this rich dataset free from spurious nuclei or DEGs. While the number of DEGs found here is significantly lower, much greater confidence can be had that these are AD-relevant genes. The low number of DEGs found may also cause concern given the sample size and cost of collection and sequencing of such datasets. However, the increasing number of snRNA-seq studies being conducted for AD creates the opportunity to conduct differential meta-analyses to increase power. Further work is required in the field to develop methods to conduct such analysis, integrating studies and accounting for their heterogeneity, similar to that which has been done for bulk RNA-seq (Rau et al., 2014). Some such approaches have already been made in COVID-19 research which could be leveraged for neurodegenerative disease (Garg et al., 2021).

The differentially expressed genes and processed count matrix from the original study are available with their manuscript. The count matrix and metadata from our reprocessing approach are available via the AD Knowledge Portal (https://adknowledgeportal.org). The AD Knowledge Portal is a platform for accessing data, analyses, and tools generated by the Accelerating Medicines Partnership (AMP-AD) Target Discovery Program and other National Institute on Aging (NIA)-supported programs to enable open-science practices and accelerate translational learning. The data, analyses, and tools are shared early in the research cycle without a publication embargo on secondary use. Data is available for general research use according to the following requirements for data access and data attribution (https://adknowledgeportal.org/DataAccess/Instructions). For access to content described in this article, see https://doi.org/10.7303/syn51758062.1. All other relevant scripts and data for working with this dataset and supporting the key findings of this study are available within the article or from our GitHub repository (copy archived at Murphy, 2023).

1. 1. Mathys H
   2. Davila-Velderrain J
   3. Peng Z
   4. Gao F
   5. Mohammadi S
   6. Young JZ
   7. Menon M
   8. He L
   9. Abdurrob F
   10. Jiang X
   11. Martorell AJ
   12. Ransohoff RM
   13. Hafler BP
   14. Bennett DA
   15. Kellis M
   16. Tsai LH

   (2019) Synapse

   Single-cell transcriptomic analysis of Alzheimer's disease.

   https://doi.org/10.7303/syn18485175

1. 1. Amezquita RA
   2. Lun ATL
   3. Becht E
   4. Carey VJ
   5. Carpp LN
   6. Geistlinger L
   7. Marini F
   8. Rue-Albrecht K
   9. Risso D
   10. Soneson C
   11. Waldron L
   12. Pagès H
   13. Smith ML
   14. Huber W
   15. Morgan M
   16. Gottardo R
   17. Hicks SC

   (2020) Publisher correction: Orchestrating single-cell analysis with bioconductor

   Nature Methods 17:137–145.

   https://doi.org/10.1038/s41592-019-0700-8

   • PubMed
   • Google Scholar
2. 1. Brase L
   2. You SF
   3. D’Oliveira Albanus R
   4. Del-Aguila JL
   5. Dai Y
   6. Novotny BC
   7. Soriano-Tarraga C
   8. Dykstra T
   9. Fernandez MV
   10. Budde JP
   11. Bergmann K
   12. Morris JC
   13. Bateman RJ
   14. Perrin RJ
   15. McDade E
   16. Xiong C
   17. Goate AM
   18. Farlow M
   19. Dominantly Inherited Alzheimer Network (DIAN)
   20. Sutherland GT
   21. Kipnis J
   22. Karch CM
   23. Benitez BA
   24. Harari O

   (2023) Single-nucleus RNA-sequencing of autosomal dominant Alzheimer disease and risk variant carriers

   Nature Communications 14:2314.

   https://doi.org/10.1038/s41467-023-37437-5

   • PubMed
   • Google Scholar
3. 1. Chen Y
   2. Lun ATL
   3. Smyth GK

   (2016) From reads to genes to pathways: differential expression analysis of RNA-Seq experiments using Rsubread and the edgeR quasi-likelihood pipeline

   F1000Research 5:1438.

   https://doi.org/10.12688/f1000research.8987.2

   • PubMed
   • Google Scholar
4. 1. Crowell HL
   2. Soneson C
   3. Germain P-L
   4. Calini D
   5. Collin L
   6. Raposo C
   7. Malhotra D
   8. Robinson MD

   (2020) muscat detects subpopulation-specific state transitions from multi-sample multi-condition single-cell transcriptomics data

   Nature Communications 11:6077.

   https://doi.org/10.1038/s41467-020-19894-4

   • PubMed
   • Google Scholar
5. 1. Fernandes HJR
   2. Patikas N
   3. Foskolou S
   4. Field SF
   5. Park J-E
   6. Byrne ML
   7. Bassett AR
   8. Metzakopian E

   (2020) Single-cell transcriptomics of parkinson’s disease human in vitro models reveals dopamine neuron-specific stress responses

   Cell Reports 33:108263.

   https://doi.org/10.1016/j.celrep.2020.108263

   • PubMed
   • Google Scholar
6. 1. Garg M
   2. Li X
   3. Moreno P
   4. Papatheodorou I
   5. Shu Y
   6. Brazma A
   7. Miao Z

   (2021) Meta-analysis of COVID-19 single-cell studies confirms eight key immune responses

   Scientific Reports 11:20833.

   https://doi.org/10.1038/s41598-021-00121-z

   • PubMed
   • Google Scholar
7. 1. Hao Y
   2. Hao S
   3. Andersen-Nissen E
   4. Mauck WM
   5. Zheng S
   6. Butler A
   7. Lee MJ
   8. Wilk AJ
   9. Darby C
   10. Zager M
   11. Hoffman P
   12. Stoeckius M
   13. Papalexi E
   14. Mimitou EP
   15. Jain J
   16. Srivastava A
   17. Stuart T
   18. Fleming LM
   19. Yeung B
   20. Rogers AJ
   21. McElrath JM
   22. Blish CA
   23. Gottardo R
   24. Smibert P
   25. Satija R

   (2021) Integrated analysis of multimodal single-cell data

   Cell 184:3573–3587.

   https://doi.org/10.1016/j.cell.2021.04.048

   • PubMed
   • Google Scholar
8. 1. Heumos L
   2. Schaar AC
   3. Lance C
   4. Litinetskaya A
   5. Drost F
   6. Zappia L
   7. Lücken MD
   8. Strobl DC
   9. Henao J
   10. Curion F
   11. Single-cell Best Practices Consortium
   12. Schiller HB
   13. Theis FJ

   (2023) Best practices for single-cell analysis across modalities

   Nature Reviews. Genetics 24:550–572.

   https://doi.org/10.1038/s41576-023-00586-w

   • PubMed
   • Google Scholar
9. 1. Hodge RD
   2. Bakken TE
   3. Miller JA
   4. Smith KA
   5. Barkan ER
   6. Graybuck LT
   7. Close JL
   8. Long B
   9. Johansen N
   10. Penn O
   11. Yao Z
   12. Eggermont J
   13. Höllt T
   14. Levi BP
   15. Shehata SI
   16. Aevermann B
   17. Beller A
   18. Bertagnolli D
   19. Brouner K
   20. Casper T
   21. Cobbs C
   22. Dalley R
   23. Dee N
   24. Ding S-L
   25. Ellenbogen RG
   26. Fong O
   27. Garren E
   28. Goldy J
   29. Gwinn RP
   30. Hirschstein D
   31. Keene CD
   32. Keshk M
   33. Ko AL
   34. Lathia K
   35. Mahfouz A
   36. Maltzer Z
   37. McGraw M
   38. Nguyen TN
   39. Nyhus J
   40. Ojemann JG
   41. Oldre A
   42. Parry S
   43. Reynolds S
   44. Rimorin C
   45. Shapovalova NV
   46. Somasundaram S
   47. Szafer A
   48. Thomsen ER
   49. Tieu M
   50. Quon G
   51. Scheuermann RH
   52. Yuste R
   53. Sunkin SM
   54. Lelieveldt B
   55. Feng D
   56. Ng L
   57. Bernard A
   58. Hawrylycz M
   59. Phillips JW
   60. Tasic B
   61. Zeng H
   62. Jones AR
   63. Koch C
   64. Lein ES

   (2019) Conserved cell types with divergent features in human versus mouse cortex

   Nature 573:61–68.

   https://doi.org/10.1038/s41586-019-1506-7

   • PubMed
   • Google Scholar
10. 1. Ilicic T
    2. Kim JK
    3. Kolodziejczyk AA
    4. Bagger FO
    5. McCarthy DJ
    6. Marioni JC
    7. Teichmann SA

    (2016) Classification of low quality cells from single-cell RNA-seq data

    Genome Biology 17:29.

    https://doi.org/10.1186/s13059-016-0888-1

    • PubMed
    • Google Scholar
11. Preprint

    1. Khozoie C
    2. Fancy N
    3. Marjaneh MM
    4. Murphy AE
    5. Matthews PM
    6. Skene N

    (2021) scFlow: A Scalable and Reproducible Analysis Pipeline for Single-Cell RNA Sequencing Data

    bioRxiv.

    https://doi.org/10.1101/2021.08.16.456499

    • Google Scholar
12. 1. Lazic SE

    (2010) The problem of pseudoreplication in neuroscientific studies: is it affecting your analysis?

    BMC Neuroscience 11:5.

    https://doi.org/10.1186/1471-2202-11-5

    • PubMed
    • Google Scholar
13. 1. Lui JH
    2. Nguyen ND
    3. Grutzner SM
    4. Darmanis S
    5. Peixoto D
    6. Wagner MJ
    7. Allen WE
    8. Kebschull JM
    9. Richman EB
    10. Ren J
    11. Newsome WT
    12. Quake SR
    13. Luo L

    (2021) Differential encoding in prefrontal cortex projection neuron classes across cognitive tasks

    Cell 184:489–506.

    https://doi.org/10.1016/j.cell.2020.11.046

    • PubMed
    • Google Scholar
14. 1. Lun ATL
    2. Riesenfeld S
    3. Andrews T
    4. Dao TP
    5. Gomes T
    6. participants in the 1st Human Cell Atlas Jamboree
    7. Marioni JC

    (2019) EmptyDrops: distinguishing cells from empty droplets in droplet-based single-cell RNA sequencing data

    Genome Biology 20:63.

    https://doi.org/10.1186/s13059-019-1662-y

    • PubMed
    • Google Scholar
15. 1. Mathys H
    2. Davila-Velderrain J
    3. Peng Z
    4. Gao F
    5. Mohammadi S
    6. Young JZ
    7. Menon M
    8. He L
    9. Abdurrob F
    10. Jiang X
    11. Martorell AJ
    12. Ransohoff RM
    13. Hafler BP
    14. Bennett DA
    15. Kellis M
    16. Tsai L-H

    (2019) Author Correction: Single-cell transcriptomic analysis of Alzheimer’s disease

    Nature 571:332–337.

    https://doi.org/10.1038/s41586-019-1329-6

    • PubMed
    • Google Scholar
16. 1. McGinnis CS
    2. Murrow LM
    3. Gartner ZJ

    (2019) DoubletFinder: doublet detection in Single-Cell RNA Sequencing data using artificial nearest Neighbors

    Cell Systems 8:329–337.

    https://doi.org/10.1016/j.cels.2019.03.003

    • PubMed
    • Google Scholar
17. Preprint

    1. McInnes L
    2. Healy J
    3. Melville J

    (2020) UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction

    arXiv.

    https://doi.org/10.48550/arXiv.1802.03426

    • Google Scholar
18. 1. McQuade A
    2. Blurton-Jones M

    (2019) Microglia in alzheimer’s disease: Exploring how genetics and phenotype influence risk

    Journal of Molecular Biology 431:1805–1817.

    https://doi.org/10.1016/j.jmb.2019.01.045

    • PubMed
    • Google Scholar
19. 1. Murphy AE
    2. Skene NG

    (2022) A balanced measure shows superior performance of pseudobulk methods in single-cell RNA-sequencing analysis

    Nature Communications 13:7851.

    https://doi.org/10.1038/s41467-022-35519-4

    • PubMed
    • Google Scholar
20. Software

    1. Murphy A

    (2023) Reanalysis_Mathys_2019, version swh:1:rev:62ffd3e446857cbdc16b59c22349ac9073d31fd0

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:25b7acdf9e62f0be7917ae981dbbcb2294196674;origin=https://github.com/neurogenomics/reanalysis_Mathys_2019;visit=swh:1:snp:baea453ad01b4118a5bb381b729fd1a1f3e4e356;anchor=swh:1:rev:62ffd3e446857cbdc16b59c22349ac9073d31fd0
21. 1. Rau A
    2. Marot G
    3. Jaffrézic F

    (2014) Differential meta-analysis of RNA-seq data from multiple studies

    BMC Bioinformatics 15:91.

    https://doi.org/10.1186/1471-2105-15-91

    • PubMed
    • Google Scholar
22. 1. Skene NG
    2. Grant SGN

    (2016) Identification of vulnerable cell types in major brain disorders using single cell transcriptomes and expression weighted cell type enrichment

    Frontiers in Neuroscience 10:16.

    https://doi.org/10.3389/fnins.2016.00016

    • PubMed
    • Google Scholar
23. 1. Soneson C
    2. Robinson MD

    (2018) Bias, robustness and scalability in single-cell differential expression analysis

    Nature Methods 15:255–261.

    https://doi.org/10.1038/nmeth.4612

    • PubMed
    • Google Scholar
24. 1. Squair JW
    2. Gautier M
    3. Kathe C
    4. Anderson MA
    5. James ND
    6. Hutson TH
    7. Hudelle R
    8. Qaiser T
    9. Matson KJE
    10. Barraud Q
    11. Levine AJ
    12. La Manno G
    13. Skinnider MA
    14. Courtine G

    (2021) Confronting false discoveries in single-cell differential expression

    Nature Communications 12:5692.

    https://doi.org/10.1038/s41467-021-25960-2

    • PubMed
    • Google Scholar
25. 1. Tasic B
    2. Yao Z
    3. Graybuck LT
    4. Smith KA
    5. Nguyen TN
    6. Bertagnolli D
    7. Goldy J
    8. Garren E
    9. Economo MN
    10. Viswanathan S
    11. Penn O
    12. Bakken T
    13. Menon V
    14. Miller J
    15. Fong O
    16. Hirokawa KE
    17. Lathia K
    18. Rimorin C
    19. Tieu M
    20. Larsen R
    21. Casper T
    22. Barkan E
    23. Kroll M
    24. Parry S
    25. Shapovalova NV
    26. Hirschstein D
    27. Pendergraft J
    28. Sullivan HA
    29. Kim TK
    30. Szafer A
    31. Dee N
    32. Groblewski P
    33. Wickersham I
    34. Cetin A
    35. Harris JA
    36. Levi BP
    37. Sunkin SM
    38. Madisen L
    39. Daigle TL
    40. Looger L
    41. Bernard A
    42. Phillips J
    43. Lein E
    44. Hawrylycz M
    45. Svoboda K
    46. Jones AR
    47. Koch C
    48. Zeng H

    (2018) Shared and distinct transcriptomic cell types across neocortical areas

    Nature 563:72–78.

    https://doi.org/10.1038/s41586-018-0654-5

    • PubMed
    • Google Scholar
26. 1. Traag VA
    2. Waltman L
    3. van Eck NJ

    (2019) From Louvain to Leiden: guaranteeing well-connected communities

    Scientific Reports 9:5233.

    https://doi.org/10.1038/s41598-019-41695-z

    • PubMed
    • Google Scholar
27. 1. Tran HTN
    2. Ang KS
    3. Chevrier M
    4. Zhang X
    5. Lee NYS
    6. Goh M
    7. Chen J

    (2020) A benchmark of batch-effect correction methods for single-cell RNA sequencing data

    Genome Biology 21:12.

    https://doi.org/10.1186/s13059-019-1850-9

    • PubMed
    • Google Scholar
28. 1. Trapnell C
    2. Cacchiarelli D
    3. Grimsby J
    4. Pokharel P
    5. Li S
    6. Morse M
    7. Lennon NJ
    8. Livak KJ
    9. Mikkelsen TS
    10. Rinn JL

    (2014) The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells

    Nature Biotechnology 32:381–386.

    https://doi.org/10.1038/nbt.2859

    • PubMed
    • Google Scholar
29. 1. Wakhloo D
    2. Scharkowski F
    3. Curto Y
    4. Javed Butt U
    5. Bansal V
    6. Steixner-Kumar AA
    7. Wüstefeld L
    8. Rajput A
    9. Arinrad S
    10. Zillmann MR
    11. Seelbach A
    12. Hassouna I
    13. Schneider K
    14. Qadir Ibrahim A
    15. Werner HB
    16. Martens H
    17. Miskowiak K
    18. Wojcik SM
    19. Bonn S
    20. Nacher J
    21. Nave K-A
    22. Ehrenreich H

    (2020) Functional hypoxia drives neuroplasticity and neurogenesis via brain erythropoietin

    Nature Communications 11:1313.

    https://doi.org/10.1038/s41467-020-15041-1

    • PubMed
    • Google Scholar
30. 1. Welch JD
    2. Kozareva V
    3. Ferreira A
    4. Vanderburg C
    5. Martin C
    6. Macosko EZ

    (2019) Single-cell multi-omic integration compares and contrasts features of brain cell identity

    Cell 177:1873–1887.

    https://doi.org/10.1016/j.cell.2019.05.006

    • PubMed
    • Google Scholar
31. 1. Xi NM
    2. Li JJ

    (2021) Benchmarking computational doublet-detection methods for Single-Cell RNA Sequencing Data

    Cell Systems 12:176–194.

    https://doi.org/10.1016/j.cels.2020.11.008

    • PubMed
    • Google Scholar
32. 1. Zimmerman KD
    2. Espeland MA
    3. Langefeld CD

    (2021) A practical solution to pseudoreplication bias in single-cell studies

    Nature Communications 12:738.

    https://doi.org/10.1038/s41467-021-21038-1

    • PubMed
    • Google Scholar

Author details

1. Alan E Murphy

   1. UK Dementia Research Institute at Imperial College London, London, United Kingdom
   2. Department of Brain Sciences, Imperial College London, London, United Kingdom

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft

   For correspondence

   alanmurph94@hotmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-2487-8753

2. Nurun Fancy

   1. UK Dementia Research Institute at Imperial College London, London, United Kingdom
   2. Department of Brain Sciences, Imperial College London, London, United Kingdom

   Contribution

   Data curation, Software, Formal analysis, Validation

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6481-6266

3. Nathan Skene

   1. UK Dementia Research Institute at Imperial College London, London, United Kingdom
   2. Department of Brain Sciences, Imperial College London, London, United Kingdom

   Contribution

   Conceptualization, Supervision, Methodology, Project administration, Writing - review and editing

   For correspondence

   n.skene@imperial.ac.uk

   Competing interests

   No competing interests declared

• 2,662

  views

• 165

  downloads

• 7

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.