New protein-coding genes emerge over time, either through a combination of rearrangement, duplication, and divergence from ancestrally coding sequences, or de novo from previously non-coding DNA (for a review see Van Oss and Carvunis, 2019, Long et al., 2003). Although genes clearly originated de novo during the emergence of life, de novo gene birth after this unique event was once believed to be extremely rare (Jacob, 1977; Keese and Gibbs, 1992; Zuckerkandl, 1975). Today, de novo gene birth is increasingly acknowledged to be real and important, with well-documented examples confirmed in diverse taxa (Baalsrud et al., 2018; Cai et al., 2008; Khalturin et al., 2009; Milde et al., 2009; Zile et al., 2020), despite the technical challenges associated with correctly identifying de novo genes (McLysaght and Hurst, 2016). Compared to genes born early in the history of life, recently emerged de novo genes have had little time to evolve and adapt. If their properties converge only slowly to those of ancient genes, this creates trends in gene properties with age (Neme and Tautz, 2013).

It has been claimed that younger genes encode shorter and faster evolving proteins (Lipman et al., 2002; Toll-Riera et al., 2012) that are more disordered, containing fewer hydrophobic amino acids (Wilson et al., 2017), which are more clustered together along the primary sequence (Foy et al., 2019). Foy et al., 2019 suggest that the two latter trends, in hydrophobicity and in its clustering, together indicate that evolution is slow to find the sophisticated folding strategies employed by older proteins, which minimize their propensity to form aggregates within cells despite relatively high overall hydrophobicity.

The universality of some of these trends has, however, been questioned. In particular, some authors have found that in some taxa, such as Saccharomyces cerevisiae (Carvunis et al., 2012; Vakirlis et al., 2018), young genes are in fact less disordered than older genes. Random sequences with high %GC tend to encode polypeptides with higher intrinsic structural disorder (ISD) (Ángyán et al., 2012). Differences in genomic GC content between species might therefore shape disorder at the time of gene birth differently (Basile et al., 2016; Van Oss and Carvunis, 2019). Because GC content evolves over relatively short timescales, GC content could explain differences in de novo genes among species, but less so among broad taxonomic groups.

Alternatively, the low disorder of young ‘protogenes’ may be an artifact of including sequences that do not meet the evolutionary definition of functionality, which requires that the loss of the gene be deleterious, that is, have a selection coefficient of less than zero (Graur et al., 2013). In a species whose %GC gives rise to low structural disorder in non-functional protogenes, the pooling of high-disorder functional and low-disorder non-functional genes in varying proportions can create a spurious trend with age. Indeed, the direction of the disorder trend in S. cerevisiae was reversed when dubious gene candidates were removed from the data of Carvunis et al., 2012 (Wilson et al., 2017).

Nevertheless, with phylostratigraphy so far studying a limited set of relatively well-annotated focal species, it remains possible that trends in protein properties as a function of age are not universal, and instead depend on taxonomic group. The forces that shape gene birth could differ across lineages because the first imperative to ‘do no harm’ operates differently as a function of differences in proteostasis machinery. Alternatively, different amino acids might be abundant, and hence cheap to use, in different taxa. We therefore set out to test whether trends in protein properties are the same vs. different across sequences of different ages and across different species, and which forces might underlie any differences.

Identifying phylostratigraphy trends as a function of gene age requires age estimation. Gene ages are based on the date of the most basal node in the phylogeny of lineages containing homologs (Domazet-Loso et al., 2007). But when sequences are highly divergent, programs used to detect homologs such as BLASTp (Altschul et al., 1990) are prone to false negatives, and thus underestimate gene age (Elhaik et al., 2006; Jain et al., 2019; McLysaght and Hurst, 2016; Moyers and Zhang, 2015; Wolfe, 2004). This is particularly problematic when studying protein properties such as length, evolutionary rate, and degree of conserved structure, because these properties themselves directly impact our ability to detect sequence similarity. Simulation studies have shown that these biases in homology detection have the potential to drive spurious phylostratigraphic trends (Elhaik et al., 2006; Moyers and Zhang, 2015; Moyers and Zhang, 2016). However, the number of ‘true’ young genes identified by phylostratigraphy studies may dwarf the numbers of genes that are falsely identified as young due to errors in homology detection, making that bias insufficient to explain previously observed trends (Domazet-Lošo et al., 2017). Statistically correcting for length and evolutionary rate did not remove observed trends in ISD (Wilson et al., 2017) or clustering (Foy et al., 2019).

Simulation studies suggest that reducing false negatives will reduce the strength of trends driven by homology detection bias (Moyers and Zhang, 2018). Therefore, if we improve the sensitivity of homology detection methods, this is predicted to result in weaker trends in protein properties with age if previously reported trends in protein properties, such as disorder and clustering, were spurious. In contrast, if previously reported trends are not artifacts, we predict that with increased sensitivity they will get stronger. This is because many false negatives may be random errors, which reduce power in detecting trends (McLysaght and Hurst, 2016), rather than systematic errors, which can create spurious trends.

Whatever the issues with homology detection, abandoning it is not a viable option. Because homologs share an evolutionary history, they are not statistically independent. But many –omic studies treat genes as independent datapoints and are thus flawed due to pseudoreplication. Phylostratigraphy can solve this problem by making each datapoint represent the independent evolutionary origin of a protein sequence, avoiding phylogenetic confounding (Felsenstein, 1985; Thornton and DeSalle, 2000). It is therefore critical that we improve upon the BLAST-based homology detection methods used within phylostratigraphy, rather than abandon the effort.

Fortunately, more recently developed methods, such as PSI-BLAST (Altschul and Koonin, 1998) and HMMER3 (Finn et al., 2011), are more sensitive to distant homology, while still being computationally efficient. Both are multiple sequence methods, which have long been known to be more sensitive to distant homologs than pairwise methods (Park et al., 1998). PSI-BLAST is a heuristic method that constructs a position-specific score matrix from the alignment of known homologs, which can then be used to iteratively search a protein database for additional homologs. HMMER instead constructs a profile HMM – a probabilistic model of the alignment – and is thus based on a formal statistical framework, yielding superior treatment of indels (Eddy, 2011). Profile HMM based methods perform significantly better than sequence based methods in comparative studies on diverse types of sequence data, with little difference in the rate of false positives (Freyhult et al., 2010; Madera and Gough, 2002).

Unfortunately, because both methods are based on iterative searches, both are vulnerable to model corruption, where a single false positive hit to a non-homologous sequence will pull in many more non-homologous sequences, potentially snowballing over multiple iterations (Pearson et al., 2017). Automated pipelines for whole-genome analysis are technically challenging, and HMMER is generally used with manual supervision of alignments.

Here we increase the sensitivity of homology detection during phylostratigraphy by using the manually curated pfam database, which was constructed using a HMMER3 pipeline. For each pfam, an HMM was built from high quality seed alignments, homologs were then found, and matches above a curated threshold were realigned back to the HMM (El-Gebali et al., 2019).

Pfams are intended to correspond to protein domains, which are structural units, capable of folding independently (Holm and Sander, 1994; for a review discussing domain definitions and identification, see Ponting and Russell, 2002). These domains are often considered to be the ‘true’ units of homology, with full proteins made up of one or more domains in a variety of different combinations (Bagowski et al., 2010; Chothia et al., 2003; Koonin et al., 2000; Moore et al., 2008). The pfam database has developed over time and now curates sets of clearly homologous sequences that are no longer required to correspond to compact folds. We continue, however, to use the term ‘domain’ in conjunction with pfams.

Here we apply HMMER3-based phylostratigraphy to a broad range of taxa. A key methodological innovation is to use pfam domains rather than BLASTp as our unit of homology. A second innovation is that instead of studying only sequences that are present in one or a few focal species, and using other species only to assign age of origin, we make full use of 435 fully sequenced species, effectively treating most of them as focal species. This not only increases our power, but also lets us evaluate how general trends are across different phylogenetic groups. We use only those pfams that are present in at least two species, a quality filter that ensures the exclusion of non-genic contaminants from our data set. We also analyze patterns of protein evolution in full genes, dating these by the oldest pfams that they contain, making results comparable to earlier phylostratigraphies based on the oldest short segment BLASTp hits (Albà and Castresana, 2005; Domazet-Loso et al., 2007). Additionally, to avoid spurious trends most likely to be due to homology detection bias, we avoid the protein metrics that most strongly affect homology detection (i.e. evolutionary rate and length), and focus on trends in properties expected a priori to have less effect on homology detection, such as the hydrophobicity and the degree of clustering of hydrophobic amino acid residues.

Our large data set gives us unprecedented power to answer two key questions. First, which trends in long-term evolution are supported rather than being due to artifacts? Here our increased power enables us to detect finer-scale trends, such as changes in the frequencies of each of the 20 amino acids, rather than simply summary statistics such as ISD. Trends in ISD and clustering get stronger with improved methods, consistent with being biological rather than due to artifacts. Second, are trends in long-term evolution universal across our tree, or specific to a particular group? To answer this, we compare trends among pfams that were born in animals, trends among pfams that were born in plants, and trends among pfams that were born prior to the origin of eukaryotes. We find that while trends in clustering are general across our data set, ISD and amino acid composition trends are specific to particular lineages – the latter compatible with specific biology rather than general artifacts of homology detection.

In this study, we analyzed all protein-coding genes containing at least one pfam annotated in the genomes of 435 species, with a total of 8209 pfams identified (see Materials and methods). We used a variety of quality filters to exclude pfams that may be due to contaminants, annotation errors or horizontal gene transfer (see Materials and methods). Eukaryotic species were included in our data set if they were marked ‘Complete’ by GOLD (Mukherjee et al., 2019), and also present in TimeTree (Hedges et al., 2006), a phylogenetic database that we used to assign evolutionary ages to pfams. A phylogenetic tree summarizing the species used in this analysis is given in Figure 1—figure supplement 1.

Where not otherwise indicated, throughout our main results we focus on non-transmembrane pfams. It is a priori likely that transmembrane pfams experience different selective pressures as they evolve, given the very different biophysical characteristics of a lipid bilayer vs. the cytosol. Pooling is therefore not appropriate unless similarities have first been established.

To avoid pseudoreplication caused by phylogenetic confounding, we take the average of each protein property across all homologs of a pfam, and then treat each homologous set as a single data point in subsequent analysis (see Materials and methods). Phylostratigraphy assigns each such set of homologous pfams to an age class, dated using timetree. We then use linear regression to obtain slope, which provides our estimate of the effect size of the relationship between sequence properties and age. In order to compare our results with those of previous studies, we also conducted our analyses on full genes.

Trends in ISD

Young genes (Willis and Masel, 2018; Wilson et al., 2017; Foy et al., 2019; Mukherjee et al., 2015) and domains (Bornberg-Bauer and Albà, 2013; Buljan and Bateman, 2009; Ekman and Elofsson, 2010; Moore and Bornberg-Bauer, 2012) have been reported to have high ISD, although some have claimed this depends on taxon (Vakirlis et al., 2018). We use IUPred to estimate ISD from amino acid sequence alone, allowing estimation across large-scale genomic data sets that contain many sequences of undetermined structure (Mészáros et al., 2018). We confirm that across our entire data set, ISD is higher in young pfams (Figure 1, linear model: R² = 0.13, p = 3 × 10⁻²⁴⁵) and genes (Figure 1—figure supplement 2A, linear model for all genes: R² = 0.057, p = 1 × 10⁻²²⁹).

Figure 1 with 4 supplements see all

Download asset Open asset

Our improved methodology increased the steepness (i.e. effect size) of the relationship between gene ISD and gene age in mouse genes above that previously estimated by Foy et al., 2019 (Figure 1—figure supplement 2B), from −0.028 to −0.050; note that mouse genes have a steeper slope than our results across all taxa. However, we note that age explains a relatively small proportion of the variance in ISD, with the remainder presumably driven by a combination of function, random biological variation, and measurement error (IUPred being an imperfect proxy for ISD or whatever other correlated biophysical property truly underlies the trend).

The slope of the relationship between ISD and age is 1.8-fold steeper for pfams (Figure 1A) than for full genes (Figure 1—figure supplement 2A); this is unsurprising, because whole genes are made up of combinations of sequences of different ages. The fact that protein domains are a more fundamental evolutionary unit than genes (Bornberg-Bauer et al., 2005; Moore et al., 2008), which is reflected in stronger relationships with pfam age, demonstrates an advantage of using a pfam-based phylostratigraphy approach. While this study focuses on pfam domains, we note that results are similar, but with smaller effect sizes, when calculated over genes.

To investigate how the ISD trend varies over time, we recalculated phylostratigraphy slopes over age-restricted subsets of our data. We also compared animal vs. plant lineages, given that these are the two kingdoms for which we have the most data (343 animal genomes and 87 plant genomes). The trend in ISD is not consistent (Figure 1B, Figure 1—figure supplement 3), but is instead driven by recent animal pfams (in which we include all pfams that emerged after the divergence of animals/fungi from plants, 1496 MYA, that are now present in animals).

There is no significant change in ISD over ‘ancient’ pfams (those that emerged prior to the last eukaryotic common ancestor [LECA], which is estimated to have existed around 2101 MYA). This does not mean that ISD is static; for example, the same domain in more exquisitely adapted vertebrates has higher ISD (Weibel, 2020).

There is also no change in ISD with age over recent plant pfams (i.e. all pfams found in plants that are younger than 1496 MY old). Note that we have relatively few plant-specific pfams in our data set compared to animal-specific pfams (Figure 1—figure supplement 3). This is likely due to differences in genome quality and annotation between the groups, and the corresponding lack of available plant genomes that meet our quality standards. However, it is also possible that these numbers reflect a biological reality that animal proteomes are richer in domains (Ramírez-Sánchez et al., 2016). We also note that animal- and plant-specific pfams have different age distributions due to differences in the respective phylogenies, and thus the set of possible divergence times in these groups.

Recent animal pfams have higher mean disorder (0.36) than recent plant pfams (0.26) (Welch’s t-test p = 4 × 10⁻⁵, figure 1a shows how this result depends on age within the recent pfam categories), with the latter still higher than ancient pfams (0.21). The difference between animals and plants does not reflect differences in the birth process alone; even in ancient pfams that are shared by plants and animals, mean ISD is 0.27 in animals vs. 0.24 in plants, a smaller but still significant difference (Wilcoxon signed rank test [a paired, non-parametric test] on difference p = 0.005). These results are compatible with animal domains experiencing more selection for high ISD than plants; plant domains have consistently less disorder with the difference being more pronounced in young, lineage-specific domains. This may be because plants produce aggregation inhibiting molecules (Velander et al., 2017), thus reducing the risk that protein aggregation poses to plant cells (see Discussion).

Trends in amino acid frequencies

IUPred scoring of ISD primarily reflects amino acid composition, with hydrophilicity being a major determinant of disorder (Dosztányi et al., 2005b; Wilson et al., 2017). Our larger data set has sufficient power to investigate age trends in the frequencies of each of the 20 amino acids individually, rather than just trends in the single IUPred summary statistic. This can reveal which amino acids drive our ISD results, as well as other potentially interesting patterns in amino acid occurrence.

Trends in amino acid frequencies with age among ancient pfams are essentially identical whether they are assessed using only plant data or only animal data (Figure 2A, Spearman’s ρ = 0.94, p = 6 × 10⁻⁶, unweighted Pearson’s R = 0.98, p = 8 × 10⁻¹⁴). In subsequent analyses beyond Figure 2, we therefore pool data across all species in our data set to calculate the phylostratigraphy slopes among ancient pfams with higher resolution.

Figure 2 with 1 supplement see all

Download asset Open asset

In contrast, recent animal pfams and recent plant pfams show different trends in amino acid frequencies with age, with slopes of similar magnitude (Welch’s t-test on absolute slope values p = 0.3), but no correlation in value (Figure 2B, p = 0.8). Recent frequency trends are mostly unrelated to ancient trends, with no relationship for animals (Figure 2C, p = 0.8), and a weak correlation for plants that is not significant (Figure 2D, p = 0.2) and is entirely driven by cysteine.

In plants, cysteine has the steepest phylostratigraphy slope. Miseta and Csutora, 2000 previously reported that %cysteine content decreased with ‘complexity’, from mammals, to plants, to a set of single celled organisms that included both photosynthesizing and non-photosynthesizing bacteria and archaea. Over all pfams, our results agree with the findings of Miseta and Csutora, in that cysteine is slightly more common in animals (2.5%) than in plants (2.2%). But surprisingly, young, plant-specific pfams have 3.5% cysteine, as opposed to 2.6% in animal-specific pfams, and 2.1% in ancient pfams.

Photosynthesis likely creates greater oxidative stress in plants than in animals, which is likely to disrupt disulfide bonds. But in the training data used by IUPred, cysteine residues are mostly in stable disulfide bonds (Mészáros et al., 2018), so IUPred scores cysteines as highly order promoting. If unpaired cysteines are common in plants, this might obscure a trend in ISD. However, recalculating IUPred scores from sequences from which we had excised cysteine resulted in very little change in our ISD results (Figure 1—figure supplement 4).

In young animal pfams, the two amino acids that are most enriched are serine and proline. These are two of the amino acids previously identified as most responsible for differences between eukaryotes and prokaryotes, due in particular to the greater quantity of ‘linker’ regions (protein sequences between pfams) of eukaryotic proteins (Basile et al., 2019). Our results suggest instead that there is nothing special about linker regions. Linker regions might simply be young sequences that happen not to have been annotated as a pfam domain. We find that not just linkers, but also young annotated pfam domains in animals have a high percentage composition of serine and proline. This is perhaps unsurprising; the simple categorization of sequences as either ‘pfam’ or ‘linker’ in genome annotation is artificial and might introduce a variety of biases.

Because animal species with more tyrosine kinases have less tyrosine, it has been argued that selection against deleterious tyrosine phosphorylation has driven a decline in tyrosine content in metazoa (Tan et al., 2009). However, evidence that selection drives tyrosine loss is limited (Pandya et al., 2015), and it has been argued that changes in GC content might explain trends in tyrosine (Su et al., 2011, although see Tan et al., 2011). Examining this question with a phylostratigraphy approach for the first time, we do not see evidence for tyrosine loss in animals; tyrosine has a relatively shallow phylostratigraphy slope that is indistinguishable from 0 (Supplementary file 1). This agrees with the findings of Pandya et al., 2015, who found no significant difference between the frequency spectra of alleles that removed or created tyrosine, suggesting that tyrosine is neither strongly selected for nor against, with trends in tyrosine being modest compared with those of other amino acids. There is, however, more tyrosine in younger ancient pfams than in the oldest ancient pfams, compatible with an ancient rather than recent process of tyrosine loss (see Supplementary file 1).

To assess whether animal-specific trends also reflect amino acid availability, we examined essential amino acids (i.e. amino acids that plants but not animals can synthesize [Guedes et al., 2011]), expecting them to be rarer in young animal domains. While this is the case (with six of nine of their phylostratigraphy slopes being positive; Figure 2), the difference is not significant (Binomial test with 50% chance of a positive slope, one-tailed p = 0.3).

Instead, the ISD score returned by IUPred seems to best summarize animal-specific trends in overall sequence hydrophobicity. We do not find differences among amino acids in hydrophobicity scores such as relative solvent accessibility (RSA) (Tien et al., 2013). While we do see a tendency for hydrophobic amino acids to have positive phylostratigraphy slopes among animal pfams and negative slopes among plant pfams (amino acids are colored by RSA in Figures 2 and 3), this correlation between RSA and phylostratigraphy slope is not significant for any of the lineage- or phylostrata-specific subsets of the data set shown in Figure 2 (see Figure 2—figure supplement 1).

Figure 3 with 2 supplements see all

Download asset Open asset

What is more, if hydrophobicity was the main determinant of phylostratigraphy slopes, we would expect amino acid composition to evolve differently in a hydrophobic membrane environment than in the cytosol. However, amino acid slopes are highly correlated between transmembrane and non-transmembrane ancient pfams (Figure 3A). Clearly, trends among ancient domains are not primarily driven by hydrophobicity. Amino acid slopes are also weakly, albeit not significantly, correlated between transmembrane and non-transmembrane pfams in animals (Figure 3B). Breaking this animal correlation is leucine, a clear outlier in that it is enriched in young transmembrane pfams but depleted in young non-transmembrane pfams. Leucine also shows the same pattern to some degree among ancient pfams. There is very little power to detect a correlation in plants, should it exist (Figure 3C).

There has been concern that phylostratigraphy slopes could be driven by homology detection bias (Moyers and Zhang, 2015, Moyers and Zhang, 2016; Wilson et al., 2017). In this case, amino acids that are more changeable, as assessed by the changeability scores of Tourasse and Li, 2000, making homology more difficult to detect, should be over-represented in young genes. However, we find if anything the opposite: more changeable amino acids are (slightly) enriched rather than depleted among the oldest of ancient pfams (Figure 3—figure supplement 1, Spearman’s ρ = 0.50, p = 0.02 for nontransmembrane pfams, and ρ = 0.40, p = 0.08 for transmembrane ancient pfams). Our results are therefore not driven by homology detection bias even at extremely long timescales. There is no correlation between recent animal (p = 0.57) or recent plant (p = 0.7) phylostratigraphy slopes and amino acid changeability, and therefore no evidence that these results are affected by homology detection bias either. Furthermore, we note that homology detection bias is expected to create similar patterns for all taxa. The fact that the strength and even direction of amino acid trends can be different for different taxa is itself also evidence against a strong role for homology detection bias.

Given the striking similarity in ancient trends between transmembrane and non-transmembrane domains, we note that Jordan et al., 2005 claim ongoing trends related to the origin of the genetic code. Specifically, they claim that the amino acids incorporated earliest into the genetic code (Trifonov, 2000) tend to be lost, while more recently acquired amino acids tend to be gained. Jordan et al., 2005 used a parsimony-based method of ancestral sequence reconstruction to detect trends in the amino acid composition of existing protein-coding sequences over time. The method has, however, been much criticized (Goldstein and Pollock, 2006; Hurst et al., 2006; McDonald, 2006).

Our amino acid phylostratigraphy slopes do not correlate with the flux values of Jordan et al., 2005 (Figure 3—figure supplement 2). However, we find that among ancient domains, the very oldest domains are enriched in the amino acids that were hypothesized to have been recruited into the genome first, both in nontransmembrane domains (Spearman’s ρ = −0.62, p = 0.004), and in transmembrane domains (Spearman’s ρ = −0.55, p = 0.01). This suggests that amino acid availabilities at different stages of life could have affected the composition of proteins born at those stages, in an effect that persists to this day. Specifically, it suggests that amino acids recruited later into the genetic code increased in abundance only slowly, creating long-term trends in amino acid availability among different ancient origin dates.

The order of recruitment ranking is taken from Trifonov, 2000, who estimated a possible consensus chronology from the expectations of 40 amino acid ranking criteria, including such diverse factors as results from experiments into primordial conditions, amino acid complexity, and thermostability. However, 3 of the 40 criteria were based on the amino acid compositions of protein assemblages. Therefore, to ensure non-circularity, we recalculated the consensus order, excluding these criteria and calculating the mean rank, as in Trifonov, 2000 prior to their proposed second step of smoothing by a filtering procedure. This recalculation had very little effect on the order of amino acids, and our correlation in ancient nontransmembrane pfams remains significant (Figure 4A, Spearman’s ρ = −0.58, p = 0.008), although the correlation for ancient transmembrane pfams is no longer on its own robust to multiple testing (Figure 4B, Spearman’s ρ = −0.47, p = 0.04).

Figure 4 with 2 supplements see all

Download asset Open asset

The order of recruitment ranking is highly speculative, and correlations with it could be driven by a single correlated amino acid physiochemical property. Specifically, amino acids with very simple structures (A, G, and V) have steeper positive phylostratigraphy slopes than any other amino acids, and this alone may drive our observed correlation between phylostratigraphy slopes and order of recruitment. We note that order of recruitment is not significantly correlated with phylostratigraphy slopes in more recent lineages (Figure 4—figure supplement 1). This is in agreement with the broader hypothesis of Jordan et al., 2005 that the order of recruitment shaped early amino acid availabilities and hence protein compositions, but is hard to reconcile with alternative explanations in which amino acid simplicity is directly causally responsible.

Our interpretation of these results is that amino acids that were introduced late into the genetic code remained relatively rare for a prolonged period of time, well after the genetic code was complete. To determine for how long, we ask when the correlation between phylostratigraphy slope and order of recruitment is strongest (Figure 4C). The strongest correlation is with slope over the three oldest phylostrata in our data set. These span the pfams found in last universal common ancestor (LUCA), through to pfams present in all extant eukaryotic but no prokaryotic lineages. The strength and significance of the correlation between order of recruitment and phylostratigraphy slope decrease if younger phylostrata are included, suggesting that from the LECA onward, the order of recruitment into the genetic code no longer shaped amino acid availability.

We do not believe our order of recruitment results to be a byproduct of a correlation with amino acid costliness. This is because there is no more than marginal significance for correlations between phylostratigraphy slope and the metabolic cost of amino acid production in yeast (Figure 4—figure supplement 2; amino acids are color-coded by costliness in Figure 4; Raiford et al., 2008).

Trends in clustering

Finally, we consider whether there are any trends in amino acid order. Specifically, a temporal trend in the value of a ‘clustering’ metric has previously been reported for mouse gene families (Foy et al., 2019). This clustering metric calculates the degree to which hydrophobic amino acids tend to lie close together along the primary sequence, normalized to ensure that values do not depend on length or amino acid frequencies (Irbäck et al., 1996) (see Materials and methods for a full technical description). Not only is there no causal link by which amino acid composition affects clustering, there is also no significant correlation (Pearson’s p = 0.06 across non-transmembrane pfams) We confirm, using our considerably larger data set, that young pfams (Figure 5A, slope = −0.037, R² = 0.028, p = 1 × 10⁻⁵⁰), and young proteins (Figure 5—figure supplement 1A; slope = −0.031, R² = 0.033, p = 5 × 10⁻¹³²) have more clustering of their hydrophobic amino acids. In Figure 5—figure supplement 1B, we show that our improved methods for assigning gene age result in a steeper slope, that is, a larger effect size, for mouse genes than that reported by Foy et al., 2019: −0.056 instead of −0.045. We note that while the proportion of variance in clustering that is explained by age is small, this is driven at least in part by large denominator variance due by domain function or random variation.

Figure 5 with 2 supplements see all

Download asset Open asset

Unlike the trend in ISD, the trend in clustering with age is remarkably consistent across age categories (Figure 5—figure supplement 2, Figure 5B: slope = −0.067, p = 0.0005 for animal-specific domains and slope = −0.025, p = 0.008 for ancient domains). While the confidence intervals increase for subsets that contain less data (such as the set of young plant pfams, for which we have little power), they always overlap the clustering slope calculated over all lineages and all phylostrata. This suggests that trends in clustering may be universal. If new sequences were born with similar clustering values, this would imply that the dispersion of hydrophobic amino acids change at an approximately constant rate over the whole of evolutionary time, with the observed trend resulting from the fact that the change has had more time to take place in older sequences. Under this interpretation, the data is also compatible with a slower contemporary rate of change of clustering in ancient pfams.

We have shown that in our data set of eukaryote genomes, there is a universal trend for young genes and pfam domains to have clustered hydrophobic amino acid residues, while old domains and genes have more evenly interspersed hydrophobic amino acids. In contrast, trends in amino acid content depend on which lineage is examined. Only animal trends are driven by young protein-coding sequences having high ISD. Ancient trends, from LUCA to the LECA, are driven by the order of recruitment of amino acids into the genetic code. We have less power to decipher plant trends, although we note a particularly strong excess of cysteine, especially relative to the generally low levels at which abundant cellular cysteine is incorporated into plant proteins as a whole.

The effect sizes of sequence age on amino acid frequencies are small. The question is whether they are biologically important. Kosinski et al., 2020 estimated the marginal effects of substituting one amino acid on the fitness associated with expressing a random peptide in E. coli. Encouragingly, they found that those marginal effects correlate with our animal phylostratigraphy slopes, suggesting that our results are indeed driven by the forces that facilitate gene birth. For example, a single switch of one proline in a 50-mer changes genotype frequency by 3% over the course of one cycle of experimental evolution. We can think of that switch as a two percentage point change in proline frequency, that is, of the order of change we see per billion years. While this fitness effect is small, it nevertheless has the potential to be biologically relevant.

Our analysis relies on homology detection of pfam domains. It is important to consider whether systematic errors in homology detection could bias our results. Our improved hmmer-based homology detection methodology should reduce the false negative rate, and thus the scope for homology detection bias. But both our ISD and our clustering trends are stronger rather than weaker than those previously reported using blastp-based methods (Foy et al., 2019; Wilson et al., 2017). Homology detection bias also cannot explain why trends in ISD or amino acid composition are lineage-specific, nor the absence of correlation with amino acids’ evolutionary changeabilities. Additionally, old sequences are expected to be most affected by homology detection bias (Moyers and Zhang, 2017), but it is more recent animal domains that drive the ISD result. Overall, we find substantial evidence contradicting the suggestion that homology detection bias drives trends.

The rejection of homology detection bias does not imply that distant homology is always successfully detected, but rather that errors in age assignment might be primarily random rather than systematic. Evidence is accumulating that de novo gene birth is not, as once thought, rare (or even non-existent) after the origin of life. A recent analysis exploited synteny to find that most taxonomically restricted genes are not the product of divergence, but have instead emerged de novo (Vakirlis et al., 2020). Another recent study excluded genes with insufficient power to detect ancient homology, and similarly found that evidence for de novo origin persisted in a substantial number of cases (Weisman et al., 2020). To get correct trends given random failures in homology detection, all we need is for apparent ages to be correlated with true ages.

Sequences can also be assigned as older than they really are, when they are incorrectly classified as present in distantly related species (false positives), due either to overly permissive homology detection cutoffs, or to contamination (see, for example, Longo et al., 2011; Merchant et al., 2014). In addition, some domains may be present in species due to horizontal gene transfer, which is problematic for the purpose of our analysis as it means that our annotation based on presence in a species does not reflect the evolutionary history of that domain (Liebeskind et al., 2016; Ravenhall et al., 2015). We used a number of filters to remove false positives and likely products of HGT from our analysis (see Materials and methods). False positives are not expected to be subject to systematic error with respect to ISD or clustering (Moyers and Zhang, 2018). In agreement with this, we found that these steps reduced the amount of ‘noise’ in our data set, making the trends we observed stronger with each improved iteration of our quality control.

Our results are consistent with the hypothesis that selection acts to reduce the aggregation propensity of proteins. Polypeptide chains are intrinsically prone to forming aggregates that are toxic to cells (Chiti and Dobson, 2017). Young animal genes avoid aggregation by being on average more disordered and hydrophilic (Wilson et al., 2017), with their few hydrophobic residues tending to cluster together along the protein chain (Foy et al., 2019). Kosinski et al., 2020 directly showed that high ISD tends to make random de novo sequences less harmful when expressed in E. coli. However, in order to be retained, sequences must also consistently serve a function within their host cells; this often requires them to fold, which is promoted by higher hydrophobicity. It is possible that pfam domains eventually reach a stable level of disorder, a balance between selection for folding, selection against misfolding, and mutational pressure. This may be why we do not see trends in ISD in ancient domains.

Frustration between the two aims of selection might be resolved by sequences evolving more sophisticated strategies for folding while avoiding aggregation. One such strategy appears to be the ordering of amino acid residues (Bertram and Masel, 2020; Foy et al., 2019), with older, more evolved sequences having more evenly dispersed hydrophobic amino acid residues. Remarkably, we find that the trend in clustering is common to all lineages (animal, plant, and pre-LECA) included in our analysis, with no evidence for saturation at a stable level of clustering. Hydrophobic amino acids are more evenly dispersed than would be expected from chance in old proteins, ruling out a trend toward a steady state set by mutation (Foy et al., 2019), The trend lowers aggregation propensity compared to the expected value based on their amino acid compositions (Foy et al., 2019), which is compatible with the trend being due to selection having had more time to decrease the clustering of hydrophobic amino acids. If so, then even over vast amounts of evolutionary time, selection continues to reduce levels of clustering in protein domains rather than reaching a steady state. This may be because decreasing clustering is difficult. For a sequence to reduce levels of hydrophobic amino acid clustering while maintaining a similar amino acid composition, many substitutions are needed on a rugged fitness landscape (Bertram and Masel, 2020).

Disordered regions can contain ‘sticky’ amino acids at protein–protein interaction surfaces (Levy et al., 2012), whose hydrophobicity within a hydrophilic region will decrease clustering. More abundant proteins have fewer sticky residues within their disordered regions (Dubreuil et al., 2019). Older domains generally have higher protein abundance (Carvunis et al., 2012), which should cause lower stickiness and higher clustering, not the lower clustering seen in our results. Our clustering trend is therefore not driven by trends in protein ‘stickiness’, which is a possible contributor to aggregation propensity.

Protein length is another hidden factor that might contribute to the trends we see. It is a priori likely that younger proteins are shorter (Van Oss and Carvunis, 2019) with a correspondingly higher surface to volume ratio, and thus a proportionately smaller hydrophobic core and higher ISD score. However, this cannot explain our results, because while younger animal domains do tend to be shorter, shorter animal domains are actually less disordered (R² = 0.0025, p = 0.007). The reasons for this are unclear; length and disorder are, as expected from surface to volume considerations, negatively correlated in ancient domains and in recent plant domains.

Differences between taxa are surprising, with young plant domains having different trends in amino acid frequencies from young animal domains, and not sharing their preference for high ISD. Plant domains as a whole are more ordered, with ancient domains in plants also having lower ISD in plants than the same domains in animals (although the difference in ISD is most pronounced in lineage-specific domains). We speculate that plants are simply better at coping with proteostasis than other taxonomic groups, using mechanisms other than high ISD to protect themselves from protein misfolding and aggregation. Plant proteomes are rich in predicted amyloidogenic regions (Antonets and Nizhnikov, 2017a), and yet there is a paucity of evidence for aggregate formation in plants under normal conditions (Antonets and Nizhnikov, 2017b). This is perhaps due in part to the presence of molecules such as polyphenols, which inhibit aggregation (Velander et al., 2017). The effectiveness of plant molecules at inhibiting aggregation has also been related to plant longevity (Mohammad-Beigi et al., 2019). Young, ordered plant proteins may therefore be under weaker selection against aggregation than the same protein would be in an animal. Our data set does not have enough plant genomes to give us power to detect whether selection acts on clustering in plants.

The degree of cysteine enrichment in young plant domains is striking, standing out despite the low power of our plant analyses. We hypothesize that young plant domains might be cysteine-rich because cysteine is relatively available in the cell. Producing cysteine is the final step in the assimilation of sulfur by plants. Additionally, cysteine production enables the detoxification of reactive oxygen species produced during photosynthesis and is essential to chloroplast function (Bermúdez et al., 2010). In the cytosol, levels of cysteine appear to play a role in maintaining redox capacity and in pathogen resistance (Romero et al., 2014). Despite the abundance of cysteine in the cytosol, its incorporation into protein sequences, while inexpensive in terms of raw materials, might be deleterious in terms of risk of oxidative damage and unstable disulfide bonds. This might explain why cysteine is progressively lost after high abundance at the time of birth. After cysteine, the two amino acids that are most enriched in young plant pfams are glutamic acid and aspartic acid, E and D. These are the most abundant amino acids in plants as a whole (Kumar et al., 2017) and may therefore also be enriched in young plant domains due to their high availability.

Genomic GC content might determine the amino acid composition of young genes, causing high %GC taxa to spawn more hydrophilic and thus disordered young genes (Ángyán et al., 2012). Our animal vs. plant data have 50% and 47% GC, respectively (Welch’s t-test, p = 0.0001), which could account for several percentage points of difference in ISD (Ángyán et al., 2012), but not for the much larger difference in young domain ISD seen in Figure 1A. We note that GC content evolves quickly between species and is variable on a finer taxonomic scale than is considered in this study. For example, monocot and dicot plants have a greater average difference in GC content than do animals and plants (Li and Du, 2014; Romiguier et al., 2010; Šmarda et al., 2014). Differences in GC content can also not explain the relative excess of cysteine in young plant domains. From inspection of the genetic code, we expect cysteine to easily accommodate any genomic %GC between 1/3 and 2/3, and this is confirmed by its prevalence as a function of species %GC (Li et al., 2015).

Our data are consistent with a hypothesis of selection acting in the usual way, that is, that proteins born de novo go on to discover, through mutation, more sophisticated strategies for folding while avoiding aggregation and/or misfolding. A second, not mutually exclusive possibility is that the trends we observe are due to differential retention of pfam domains and genes, rather than descent with modification. Under this hypothesis, newborn sequences are diverse, and sequences with favored properties are lost less often during subsequent long-term evolution. The relative contribution of these two levels of selection could be investigated if quality control could be made stringent enough to obtain quantitatively reliable loss rates and diversification rates for pfam domains.

In conclusion, trends in the evolution of amino acid composition show surprising differences between taxonomic groups and over different spans of time. Amino acid availability at time of de novo birth has a strong effect on domain composition. Even after billions of years of evolution, ancient domains remain influenced by which amino acids were more abundant as the genetic code was being formed. De novo birth and subsequent trends in plants may also be shaped by amino acid availabilities. In young animal domains, amino acid composition trends may be due to selection for high ISD. However, the inferred ancestors of eukaryote protein sequences show a universal trend toward reducing their levels of hydrophobic clustering with age, no matter their amino acid composition, presumably in an attempt to find a balance between folding and misfolding. The only claim in evolutionary biology that is close to comparable in timescale is the increase in body size in some taxa over the 3.5 billion-year history of life (Cope, 1885; Heim et al., 2015; Payne et al., 2009).

All scripts used in this work can be accessed at: https://github.com/MaselLab/ProteinEvolution (copy archived at https://archive.softwareheritage.org/swh:1:rev:1c3b5dcc3dbce35acc1bbbfaafa29fc6398ecee8/). Our raw data, and homology files used in our analyses, are available at https://doi.org/10.6084/m9.figshare.12037281.

1. 1. James JE
   2. Willis SM
   3. Nelson PG
   4. Weibel C
   5. Kosinski LJ
   6. Masel J

   (2020) figshare

   Data from: Universal and taxon-specific trends in protein sequences as a function of age.

   https://doi.org/10.6084/m9.figshare.12037281

1. 1. Albà MM
   2. Castresana J

   (2005) Inverse relationship between evolutionary rate and age of mammalian genes

   Molecular Biology and Evolution 22:598–606.

   https://doi.org/10.1093/molbev/msi045

   • PubMed
   • Google Scholar
2. 1. Altschul SF
   2. Gish W
   3. Miller W
   4. Myers EW
   5. Lipman DJ

   (1990) Basic local alignment search tool

   Journal of Molecular Biology 215:403–410.

   https://doi.org/10.1016/S0022-2836(05)80360-2

   • PubMed
   • Google Scholar
3. 1. Altschul SF
   2. Koonin EV

   (1998) Iterated profile searches with PSI-BLAST—a tool for discovery in protein databases

   Trends in Biochemical Sciences 23:444–447.

   https://doi.org/10.1016/S0968-0004(98)01298-5

   • Google Scholar
4. 1. Ángyán AF
   2. Perczel A
   3. Gáspári Z

   (2012) Estimating intrinsic structural preferences of de novo emerging random-sequence proteins: is aggregation the main bottleneck?

   FEBS Letters 586:2468–2472.

   https://doi.org/10.1016/j.febslet.2012.06.007

   • PubMed
   • Google Scholar
5. 1. Antonets KS
   2. Nizhnikov AA

   (2017a) Amyloids and prions in plants: facts and perspectives

   Prion 11:300–312.

   https://doi.org/10.1080/19336896.2017.1377875

   • PubMed
   • Google Scholar
6. 1. Antonets K
   2. Nizhnikov A

   (2017b) Predicting amyloidogenic proteins in the proteomes of plants

   International Journal of Molecular Sciences 18:2155.

   https://doi.org/10.3390/ijms18102155

   • Google Scholar
7. 1. Baalsrud HT
   2. Tørresen OK
   3. Solbakken MH
   4. Salzburger W
   5. Hanel R
   6. Jakobsen KS
   7. Jentoft S

   (2018) De novo gene evolution of antifreeze glycoproteins in codfishes revealed by whole genome sequence data

   Molecular Biology and Evolution 35:593–606.

   https://doi.org/10.1093/molbev/msx311

   • PubMed
   • Google Scholar
8. 1. Bagowski CP
   2. Bruins W
   3. Te Velthuis AJ

   (2010) The nature of protein domain evolution: shaping the interaction network

   Current Genomics 11:368–376.

   https://doi.org/10.2174/138920210791616725

   • PubMed
   • Google Scholar
9. 1. Basile W
   2. Sachenkova O
   3. Light S
   4. Elofsson A

   (2016) High GC content causes orphan proteins to be intrinsically disordered

   PLOS Computational Biology 13:e1005375.

   https://doi.org/10.1371/journal.pcbi.1005375

   • Google Scholar
10. 1. Basile W
    2. Salvatore M
    3. Bassot C
    4. Elofsson A

    (2019) Why do eukaryotic proteins contain more intrinsically disordered regions?

    PLOS Computational Biology 15:e1007186.

    https://doi.org/10.1371/journal.pcbi.1007186

    • PubMed
    • Google Scholar
11. 1. Berkemer SJ
    2. McGlynn SE

    (2020) A new analysis of Archaea-Bacteria domain separation: variable phylogenetic distance and the tempo of early evolution

    Molecular Biology and Evolution 37:2332–2340.

    https://doi.org/10.1093/molbev/msaa089

    • PubMed
    • Google Scholar
12. 1. Bermúdez MA
    2. Páez-Ochoa MA
    3. Gotor C
    4. Romero LC

    (2010) Arabidopsis S-sulfocysteine synthase activity is essential for chloroplast function and long-day light-dependent redox control

    The Plant Cell 22:403–416.

    https://doi.org/10.1105/tpc.109.071985

    • PubMed
    • Google Scholar
13. 1. Bertram J
    2. Masel J

    (2020) Evolution rapidly optimizes stability and aggregation in lattice proteins despite pervasive landscape valleys and mazes

    Genetics 214:1047–1057.

    https://doi.org/10.1534/genetics.120.302815

    • PubMed
    • Google Scholar
14. 1. Bornberg-Bauer E
    2. Beaussart F
    3. Kummerfeld SK
    4. Teichmann SA
    5. Weiner J

    (2005) The evolution of domain arrangements in proteins and interaction networks

    Cellular and Molecular Life Sciences CMLS 62:435–445.

    https://doi.org/10.1007/s00018-004-4416-1

    • PubMed
    • Google Scholar
15. 1. Bornberg-Bauer E
    2. Albà MM

    (2013) Dynamics and adaptive benefits of modular protein evolution

    Current Opinion in Structural Biology 23:459–466.

    https://doi.org/10.1016/j.sbi.2013.02.012

    • PubMed
    • Google Scholar
16. 1. Buljan M
    2. Bateman A

    (2009) The evolution of protein domain families

    Biochemical Society Transactions 37:751–755.

    https://doi.org/10.1042/BST0370751

    • PubMed
    • Google Scholar
17. 1. Cai J
    2. Zhao R
    3. Jiang H
    4. Wang W

    (2008) De novo origination of a new protein-coding gene in Saccharomyces cerevisiae

    Genetics 179:487–496.

    https://doi.org/10.1534/genetics.107.084491

    • PubMed
    • Google Scholar
18. 1. Carvunis AR
    2. Rolland T
    3. Wapinski I
    4. Calderwood MA
    5. Yildirim MA
    6. Simonis N
    7. Charloteaux B
    8. Hidalgo CA
    9. Barbette J
    10. Santhanam B
    11. Brar GA
    12. Weissman JS
    13. Regev A
    14. Thierry-Mieg N
    15. Cusick ME
    16. Vidal M

    (2012) Proto-genes and de novo gene birth

    Nature 487:370–374.

    https://doi.org/10.1038/nature11184

    • PubMed
    • Google Scholar
19. 1. Chiti F
    2. Dobson CM

    (2017) Protein misfolding, amyloid formation, and human disease: a summary of progress over the last decade

    Annual Review of Biochemistry 86:27–68.

    https://doi.org/10.1146/annurev-biochem-061516-045115

    • PubMed
    • Google Scholar
20. 1. Chothia C
    2. Gough J
    3. Vogel C
    4. Teichmann SA

    (2003) Evolution of the protein repertoire

    Science 300:1701–1703.

    https://doi.org/10.1126/science.1085371

    • PubMed
    • Google Scholar
21. 1. Cope ED

    (1885) On the evolution of the Vertebrata, progressive and retrogressive

    The American Naturalist 19:140–148.

    https://doi.org/10.1086/273881

    • Google Scholar
22. 1. Domazet-Loso T
    2. Brajković J
    3. Tautz D

    (2007) A phylostratigraphy approach to uncover the genomic history of major adaptations in metazoan lineages

    Trends in Genetics 23:533–539.

    https://doi.org/10.1016/j.tig.2007.08.014

    • PubMed
    • Google Scholar
23. 1. Domazet-Lošo T
    2. Carvunis AR
    3. Albà MM
    4. Šestak MS
    5. Bakaric R
    6. Neme R
    7. Tautz D

    (2017) No evidence for phylostratigraphic Bias impacting inferences on patterns of gene emergence and evolution

    Molecular Biology and Evolution 34:843–856.

    https://doi.org/10.1093/molbev/msw284

    • PubMed
    • Google Scholar
24. 1. Dosztányi Z
    2. Csizmok V
    3. Tompa P
    4. Simon I

    (2005a) IUPred: web server for the prediction of intrinsically unstructured regions of proteins based on estimated energy content

    Bioinformatics 21:3433–3434.

    https://doi.org/10.1093/bioinformatics/bti541

    • PubMed
    • Google Scholar
25. 1. Dosztányi Z
    2. Csizmók V
    3. Tompa P
    4. Simon I

    (2005b) The pairwise energy content estimated from amino acid composition discriminates between folded and intrinsically unstructured proteins

    Journal of Molecular Biology 347:827–839.

    https://doi.org/10.1016/j.jmb.2005.01.071

    • PubMed
    • Google Scholar
26. 1. Dubreuil B
    2. Matalon O
    3. Levy ED

    (2019) Protein abundance biases the amino acid composition of disordered regions to minimize Non-functional interactions

    Journal of Molecular Biology 431:4978–4992.

    https://doi.org/10.1016/j.jmb.2019.08.008

    • PubMed
    • Google Scholar
27. 1. Eddy SR

    (2011) Accelerated profile HMM searches

    PLOS Computational Biology 7:e1002195.

    https://doi.org/10.1371/journal.pcbi.1002195

    • PubMed
    • Google Scholar
28. 1. Ekman D
    2. Elofsson A

    (2010) Identifying and quantifying orphan protein sequences in fungi

    Journal of Molecular Biology 396:396–405.

    https://doi.org/10.1016/j.jmb.2009.11.053

    • PubMed
    • Google Scholar
29. 1. El-Gebali S
    2. Mistry J
    3. Bateman A
    4. Eddy SR
    5. Luciani A
    6. Potter SC
    7. Qureshi M
    8. Richardson LJ
    9. Salazar GA
    10. Smart A
    11. Sonnhammer ELL
    12. Hirsh L
    13. Paladin L
    14. Piovesan D
    15. Tosatto SCE
    16. Finn RD

    (2019) The pfam protein families database in 2019

    Nucleic Acids Research 47:D427–D432.

    https://doi.org/10.1093/nar/gky995

    • Google Scholar
30. 1. Elhaik E
    2. Sabath N
    3. Graur D

    (2006) The "inverse relationship between evolutionary rate and age of mammalian genes" is an artifact of increased genetic distance with rate of evolution and time of divergence

    Molecular Biology and Evolution 23:1–3.

    https://doi.org/10.1093/molbev/msj006

    • PubMed
    • Google Scholar
31. 1. Eme L
    2. Sharpe SC
    3. Brown MW
    4. Roger AJ

    (2014) On the age of eukaryotes: evaluating evidence from fossils and molecular clocks

    Cold Spring Harbor Perspectives in Biology 6:a016139.

    https://doi.org/10.1101/cshperspect.a016139

    • PubMed
    • Google Scholar
32. 1. Felsenstein J

    (1985) Phylogenies and the comparative method

    The American Naturalist 125:1–15.

    https://doi.org/10.1086/284325

    • Google Scholar
33. 1. Finn RD
    2. Clements J
    3. Eddy SR

    (2011) HMMER web server: interactive sequence similarity searching

    Nucleic Acids Research 39:W29–W37.

    https://doi.org/10.1093/nar/gkr367

    • Google Scholar
34. 1. Fisher RA

    (1915) Frequency distribution of the values of the correlation coefficient in samples from an indefinitely large population

    Biometrika 10:507–521.

    https://doi.org/10.2307/2331838

    • Google Scholar
35. 1. Foy SG
    2. Wilson BA
    3. Bertram J
    4. Cordes MHJ
    5. Masel J

    (2019) A shift in aggregation avoidance strategy marks a Long-Term direction to protein evolution

    Genetics 211:1345–1355.

    https://doi.org/10.1534/genetics.118.301719

    • PubMed
    • Google Scholar
36. 1. Freyhult EK
    2. Bollback JP
    3. Gardner PP

    (2010) Exploring genomic dark matter: a critical assessment of the performance of homology search methods on noncoding RNA

    Genome Research 17:117–125.

    https://doi.org/10.1101/gr.5890907

    • Google Scholar
37. 1. Goldstein RA
    2. Pollock DD

    (2006) Observations of amino acid gain and loss during protein evolution are explained by statistical Bias

    Molecular Biology and Evolution 23:1444–1449.

    https://doi.org/10.1093/molbev/msl010

    • PubMed
    • Google Scholar
38. 1. Graur D
    2. Zheng Y
    3. Price N
    4. Azevedo RB
    5. Zufall RA
    6. Elhaik E

    (2013) On the immortality of television sets: "function" in the human genome according to the evolution-free gospel of ENCODE

    Genome Biology and Evolution 5:578–590.

    https://doi.org/10.1093/gbe/evt028

    • PubMed
    • Google Scholar
39. 1. Guedes RL
    2. Prosdocimi F
    3. Fernandes GR
    4. Moura LK
    5. Ribeiro HA
    6. Ortega JM

    (2011) Amino acids biosynthesis and nitrogen assimilation pathways: a great genomic deletion during eukaryotes evolution

    BMC Genomics 12:S2.

    https://doi.org/10.1186/1471-2164-12-S4-S2

    • PubMed
    • Google Scholar
40. 1. Hedges SB
    2. Chen H
    3. Kumar S
    4. Wang DY
    5. Thompson AS
    6. Watanabe H

    (2001) A genomic timescale for the origin of eukaryotes

    BMC Evolutionary Biology 1:4.

    https://doi.org/10.1186/1471-2148-1-4

    • PubMed
    • Google Scholar
41. 1. Hedges SB
    2. Dudley J
    3. Kumar S

    (2006) TimeTree: a public knowledge-base of divergence times among organisms

    Bioinformatics 22:2971–2972.

    https://doi.org/10.1093/bioinformatics/btl505

    • PubMed
    • Google Scholar
42. 1. Heim NA
    2. Knope ML
    3. Schaal EK
    4. Wang SC
    5. Payne JL

    (2015) Cope's rule in the evolution of marine animals

    Science 347:867–870.

    https://doi.org/10.1126/science.1260065

    • Google Scholar
43. 1. Holm L
    2. Sander C

    (1994) Parser for protein folding units

    Proteins: Structure, Function, and Genetics 19:256–268.

    https://doi.org/10.1002/prot.340190309

    • Google Scholar
44. 1. Hurst LD
    2. Feil EJ
    3. Rocha EPC

    (2006) Causes of trends in amino-acid gain and loss

    Nature 442:E11–E12.

    https://doi.org/10.1038/nature05137

    • Google Scholar
45. 1. Irbäck A
    2. Peterson C
    3. Potthast F

    (1996) Evidence for nonrandom hydrophobicity structures in protein chains

    PNAS 93:9533–9538.

    https://doi.org/10.1073/pnas.93.18.9533

    • PubMed
    • Google Scholar
46. 1. Jacob F

    (1977) Evolution and tinkering

    Science 196:1161–1166.

    https://doi.org/10.1126/science.860134

    • PubMed
    • Google Scholar
47. 1. Jain A
    2. Perisa D
    3. Fliedner F
    4. von Haeseler A
    5. Ebersberger I

    (2019) The evolutionary traceability of a protein

    Genome Biology and Evolution 11:531–545.

    https://doi.org/10.1093/gbe/evz008

    • PubMed
    • Google Scholar
48. 1. Jones P
    2. Binns D
    3. Chang HY
    4. Fraser M
    5. Li W
    6. McAnulla C
    7. McWilliam H
    8. Maslen J
    9. Mitchell A
    10. Nuka G
    11. Pesseat S
    12. Quinn AF
    13. Sangrador-Vegas A
    14. Scheremetjew M
    15. Yong SY
    16. Lopez R
    17. Hunter S

    (2014) InterProScan 5: genome-scale protein function classification

    Bioinformatics 30:1236–1240.

    https://doi.org/10.1093/bioinformatics/btu031

    • PubMed
    • Google Scholar
49. 1. Jordan IK
    2. Kondrashov FA
    3. Adzhubei IA
    4. Wolf YI
    5. Koonin EV
    6. Kondrashov AS
    7. Sunyaev S

    (2005) A universal trend of amino acid gain and loss in protein evolution

    Nature 433:633–638.

    https://doi.org/10.1038/nature03306

    • PubMed
    • Google Scholar
50. 1. Keese PK
    2. Gibbs A

    (1992) Origins of genes: "big bang" or continuous creation?

    PNAS 89:9489–9493.

    https://doi.org/10.1073/pnas.89.20.9489

    • PubMed
    • Google Scholar
51. 1. Khalturin K
    2. Hemmrich G
    3. Fraune S
    4. Augustin R
    5. Bosch TC

    (2009) More than just orphans: are taxonomically-restricted genes important in evolution?

    Trends in Genetics 25:404–413.

    https://doi.org/10.1016/j.tig.2009.07.006

    • PubMed
    • Google Scholar
52. 1. Kinsella RJ
    2. Kähäri A
    3. Haider S
    4. Zamora J
    5. Proctor G
    6. Spudich G
    7. Almeida-King J
    8. Staines D
    9. Derwent P
    10. Kerhornou A
    11. Kersey P
    12. Flicek P

    (2011) Ensembl BioMarts: a hub for data retrieval across taxonomic space

    Database 2011:bar030.

    https://doi.org/10.1093/database/bar030

    • PubMed
    • Google Scholar
53. 1. Koonin EV
    2. Aravind L
    3. Kondrashov AS

    (2000) The impact of comparative genomics on our understanding of evolution

    Cell 101:573–576.

    https://doi.org/10.1016/S0092-8674(00)80867-3

    • PubMed
    • Google Scholar
54. Preprint

    1. Kosinski L
    2. Aviles N
    3. Gomez K
    4. Masel J

    (2020) Amino acids that are well tolerated in random peptides in E. coli are enriched in young animal but not young plant genes

    bioRxiv.

    https://doi.org/10.1101/2020.04.28.066316

    • Google Scholar
55. 1. Krogh A
    2. Larsson B
    3. von Heijne G
    4. Sonnhammer EL

    (2001) Predicting transmembrane protein topology with a hidden Markov model: application to complete genomes

    Journal of Molecular Biology 305:567–580.

    https://doi.org/10.1006/jmbi.2000.4315

    • PubMed
    • Google Scholar
56. 1. Kryukov K
    2. Imanishi T

    (2016) Human contamination in public genome assemblies

    PLOS ONE 11:e0162424.

    https://doi.org/10.1371/journal.pone.0162424

    • PubMed
    • Google Scholar
57. 1. Kumar V
    2. Sharma A
    3. Kaur R
    4. Thukral AK
    5. Bhardwaj R
    6. Ahmad P

    (2017) Differential distribution of amino acids in plants

    Amino Acids 49:821–869.

    https://doi.org/10.1007/s00726-017-2401-x

    • PubMed
    • Google Scholar
58. 1. Letunic I
    2. Bork P

    (2006) Interactive tree of life (ITOL): An online tool for phylogenetic tree display and annotation

    Bioinformatics 21:127–128.

    https://doi.org/10.1093/bioinformatics/btl529

    • Google Scholar
59. 1. Levy ED
    2. De S
    3. Teichmann SA

    (2012) Cellular crowding imposes global constraints on the chemistry and evolution of proteomes

    PNAS 109:20461–20466.

    https://doi.org/10.1073/pnas.1209312109

    • PubMed
    • Google Scholar
60. 1. Li J
    2. Zhou J
    3. Wu Y
    4. Yang S
    5. Tian D

    (2015) GC-Content of synonymous codons profoundly influences amino acid usage

    G3: Genes, Genomes, Genetics 5:2027–2036.

    https://doi.org/10.1534/g3.115.019877

    • Google Scholar
61. 1. Li XQ
    2. Du D

    (2014) Variation, evolution, and correlation analysis of C+G content and genome or chromosome size in different kingdoms and phyla

    PLOS ONE 9:e88339.

    https://doi.org/10.1371/journal.pone.0088339

    • PubMed
    • Google Scholar
62. 1. Liebeskind BJ
    2. McWhite CD
    3. Marcotte EM

    (2016) Towards consensus gene ages

    Genome Biology and Evolution 8:1812–1823.

    https://doi.org/10.1093/gbe/evw113

    • PubMed
    • Google Scholar
63. 1. Lipman DJ
    2. Souvorov A
    3. Koonin EV
    4. Panchenko AR
    5. Tatusova TA

    (2002) The relationship of protein conservation and sequence length

    BMC Evolutionary Biology 2:20.

    https://doi.org/10.1186/1471-2148-2-20

    • PubMed
    • Google Scholar
64. 1. Long M
    2. Betrán E
    3. Thornton K
    4. Wang W

    (2003) The origin of new genes: glimpses from the young and old

    Nature Reviews Genetics 4:865–875.

    https://doi.org/10.1038/nrg1204

    • PubMed
    • Google Scholar
65. 1. Longo MS
    2. O'Neill MJ
    3. O'Neill RJ

    (2011) Abundant human DNA contamination identified in non-primate genome databases

    PLOS ONE 6:e16410.

    https://doi.org/10.1371/journal.pone.0016410

    • PubMed
    • Google Scholar
66. 1. Madera M
    2. Gough J

    (2002) A comparison of profile hidden markov model procedures for remote homology detection

    Nucleic Acids Research 30:4321–4328.

    https://doi.org/10.1093/nar/gkf544

    • PubMed
    • Google Scholar
67. 1. McDonald JH

    (2006) Apparent trends of amino acid gain and loss in protein evolution due to nearly neutral variation

    Molecular Biology and Evolution 23:240–244.

    https://doi.org/10.1093/molbev/msj026

    • PubMed
    • Google Scholar
68. 1. McLysaght A
    2. Hurst LD

    (2016) Open questions in the study of de novo genes: what, how and why

    Nature Reviews Genetics 17:567–578.

    https://doi.org/10.1038/nrg.2016.78

    • PubMed
    • Google Scholar
69. 1. Merchant S
    2. Wood DE
    3. Salzberg SL

    (2014) Unexpected cross-species contamination in genome sequencing projects

    PeerJ 2:e675.

    https://doi.org/10.7717/peerj.675

    • PubMed
    • Google Scholar
70. 1. Mészáros B
    2. Erdos G
    3. Dosztányi Z

    (2018) IUPred2A: context-dependent prediction of protein disorder as a function of redox state and protein binding

    Nucleic Acids Research 46:W329–W337.

    https://doi.org/10.1093/nar/gky384

    • PubMed
    • Google Scholar
71. 1. Milde S
    2. Hemmrich G
    3. Anton-Erxleben F
    4. Khalturin K
    5. Wittlieb J
    6. Bosch TC

    (2009) Characterization of taxonomically restricted genes in a phylum-restricted cell type

    Genome Biology 10:R8–R16.

    https://doi.org/10.1186/gb-2009-10-1-r8

    • PubMed
    • Google Scholar
72. 1. Miseta A
    2. Csutora P

    (2000) Relationship between the occurrence of cysteine in proteins and the complexity of organisms

    Molecular Biology and Evolution 17:1232–1239.

    https://doi.org/10.1093/oxfordjournals.molbev.a026406

    • PubMed
    • Google Scholar
73. 1. Mohammad-Beigi H
    2. Kjaer L
    3. Eskandari H
    4. Aliakbari F
    5. Christiansen G
    6. Ruvo G
    7. Ward JL
    8. Otzen DE

    (2019) A possible connection between plant longevity and the absence of protein fibrillation: basis for identifying aggregation inhibitors in plants

    Frontiers in Plant Science 10:148.

    https://doi.org/10.3389/fpls.2019.00148

    • PubMed
    • Google Scholar
74. 1. Moore AD
    2. Björklund AK
    3. Ekman D
    4. Bornberg-Bauer E
    5. Elofsson A

    (2008) Arrangements in the modular evolution of proteins

    Trends in Biochemical Sciences 33:444–451.

    https://doi.org/10.1016/j.tibs.2008.05.008

    • PubMed
    • Google Scholar
75. 1. Moore AD
    2. Bornberg-Bauer E

    (2012) The dynamics and evolutionary potential of domain loss and emergence

    Molecular Biology and Evolution 29:787–796.

    https://doi.org/10.1093/molbev/msr250

    • PubMed
    • Google Scholar
76. 1. Moyers BA
    2. Zhang J

    (2015) Phylostratigraphic Bias creates spurious patterns of genome evolution

    Molecular Biology and Evolution 32:258–267.

    https://doi.org/10.1093/molbev/msu286

    • PubMed
    • Google Scholar
77. 1. Moyers BA
    2. Zhang J

    (2016) Evaluating phylostratigraphic evidence for widespread de novo gene birth in genome evolution

    Molecular Biology and Evolution 33:1245–1256.

    https://doi.org/10.1093/molbev/msw008

    • PubMed
    • Google Scholar
78. 1. Moyers BA
    2. Zhang J

    (2017) Further simulations and analyses demonstrate open problems of phylostratigraphy

    Genome Biology and Evolution 9:1519–1527.

    https://doi.org/10.1093/gbe/evx109

    • PubMed
    • Google Scholar
79. 1. Moyers BA
    2. Zhang J

    (2018) Toward reducing phylostratigraphic errors and biases

    Genome Biology and Evolution 10:2037–2048.

    https://doi.org/10.1093/gbe/evy161

    • PubMed
    • Google Scholar
80. 1. Mukherjee S
    2. Panda A
    3. Ghosh TC

    (2015) Elucidating evolutionary features and functional implications of orphan genes in leishmania major

    Infection, Genetics and Evolution 32:330–337.

    https://doi.org/10.1016/j.meegid.2015.03.031

    • Google Scholar
81. 1. Mukherjee S
    2. Stamatis D
    3. Bertsch J
    4. Ovchinnikova G
    5. Katta HY
    6. Mojica A
    7. Chen IA
    8. Kyrpides NC
    9. Reddy T

    (2019) Genomes OnLine database (GOLD) v.7: updates and new features

    Nucleic Acids Research 47:D649–D659.

    https://doi.org/10.1093/nar/gky977

    • PubMed
    • Google Scholar
82. 1. Neme R
    2. Tautz D

    (2013) Phylogenetic patterns of emergence of new genes support a model of frequent de novo evolution

    BMC Genomics 14:117.

    https://doi.org/10.1186/1471-2164-14-117

    • PubMed
    • Google Scholar
83. 1. O'Leary NA
    2. Wright MW
    3. Brister JR
    4. Ciufo S
    5. Haddad D
    6. McVeigh R
    7. Rajput B
    8. Robbertse B
    9. Smith-White B
    10. Ako-Adjei D
    11. Astashyn A
    12. Badretdin A
    13. Bao Y
    14. Blinkova O
    15. Brover V
    16. Chetvernin V
    17. Choi J
    18. Cox E
    19. Ermolaeva O
    20. Farrell CM
    21. Goldfarb T
    22. Gupta T
    23. Haft D
    24. Hatcher E
    25. Hlavina W
    26. Joardar VS
    27. Kodali VK
    28. Li W
    29. Maglott D
    30. Masterson P
    31. McGarvey KM
    32. Murphy MR
    33. O'Neill K
    34. Pujar S
    35. Rangwala SH
    36. Rausch D
    37. Riddick LD
    38. Schoch C
    39. Shkeda A
    40. Storz SS
    41. Sun H
    42. Thibaud-Nissen F
    43. Tolstoy I
    44. Tully RE
    45. Vatsan AR
    46. Wallin C
    47. Webb D
    48. Wu W
    49. Landrum MJ
    50. Kimchi A
    51. Tatusova T
    52. DiCuccio M
    53. Kitts P
    54. Murphy TD
    55. Pruitt KD

    (2016) Reference sequence (RefSeq) database at NCBI: current status, taxonomic expansion, and functional annotation

    Nucleic Acids Research 44:D733–D745.

    https://doi.org/10.1093/nar/gkv1189

    • Google Scholar
84. 1. Pandya S
    2. Struck TJ
    3. Mannakee BK
    4. Paniscus M
    5. Gutenkunst RN

    (2015) Testing whether metazoan tyrosine loss was driven by selection against promiscuous phosphorylation

    Molecular Biology and Evolution 32:144–152.

    https://doi.org/10.1093/molbev/msu284

    • PubMed
    • Google Scholar
85. 1. Parfrey LW
    2. Lahr DJ
    3. Knoll AH
    4. Katz LA

    (2011) Estimating the timing of early eukaryotic diversification with multigene molecular clocks

    PNAS 108:13624–13629.

    https://doi.org/10.1073/pnas.1110633108

    • PubMed
    • Google Scholar
86. 1. Park J
    2. Karplus K
    3. Barrett C
    4. Hughey R
    5. Haussler D
    6. Hubbard T
    7. Chothia C

    (1998) Sequence comparisons using multiple sequences detect three times as many remote homologues as pairwise methods

    Journal of Molecular Biology 284:1201–1210.

    https://doi.org/10.1006/jmbi.1998.2221

    • PubMed
    • Google Scholar
87. 1. Payne JL
    2. Boyer AG
    3. Brown JH
    4. Finnegan S
    5. Kowalewski M
    6. Krause RA
    7. Lyons SK
    8. McClain CR
    9. McShea DW
    10. Novack-Gottshall PM
    11. Smith FA
    12. Stempien JA
    13. Wang SC

    (2009) Two-phase increase in the maximum size of life over 3.5 billion years reflects biological innovation and environmental opportunity

    PNAS 106:24–27.

    https://doi.org/10.1073/pnas.0806314106

    • PubMed
    • Google Scholar
88. 1. Pearson WR
    2. Li W
    3. Lopez R

    (2017) Query-seeded iterative sequence similarity searching improves selectivity 5-20-fold

    Nucleic Acids Research 45:e46.

    https://doi.org/10.1093/nar/gkw1207

    • PubMed
    • Google Scholar
89. 1. Ponting CP
    2. Russell RR

    (2002) The natural history of protein domains

    Annual Review of Biophysics and Biomolecular Structure 31:45–71.

    https://doi.org/10.1146/annurev.biophys.31.082901.134314

    • PubMed
    • Google Scholar
90. 1. Raiford DW
    2. Heizer EM
    3. Miller RV
    4. Akashi H
    5. Raymer ML
    6. Krane DE

    (2008) Do amino acid biosynthetic costs constrain protein evolution in Saccharomyces cerevisiae?

    Journal of Molecular Evolution 67:621–630.

    https://doi.org/10.1007/s00239-008-9162-9

    • PubMed
    • Google Scholar
91. 1. Ramírez-Sánchez O
    2. Pérez-Rodríguez P
    3. Delaye L
    4. Tiessen A

    (2016) Plant proteins are smaller because they are encoded by fewer exons than animal proteins

    Genomics, Proteomics & Bioinformatics 14:357–370.

    https://doi.org/10.1016/j.gpb.2016.06.003

    • Google Scholar
92. 1. Ravenhall M
    2. Škunca N
    3. Lassalle F
    4. Dessimoz C

    (2015) Inferring horizontal gene transfer

    PLOS Computational Biology 11:e1004095.

    https://doi.org/10.1371/journal.pcbi.1004095

    • PubMed
    • Google Scholar
93. Preprint

    1. Robinson D

    (2014) Broom: an R package for converting statistical analysis objects into tidy data frames

    arXiv.

    https://arxiv.org/abs/1412.3565

    • Google Scholar
94. 1. Romero LC
    2. Aroca MÁ
    3. Laureano-Marín AM
    4. Moreno I
    5. García I
    6. Gotor C

    (2014) Cysteine and cysteine-related signaling pathways in Arabidopsis thaliana

    Molecular Plant 7:264–276.

    https://doi.org/10.1093/mp/sst168

    • PubMed
    • Google Scholar
95. 1. Romiguier J
    2. Ranwez V
    3. Douzery EJ
    4. Galtier N

    (2010) Contrasting GC-content dynamics across 33 mammalian genomes: relationship with life-history traits and chromosome sizes

    Genome Research 20:1001–1009.

    https://doi.org/10.1101/gr.104372.109

    • PubMed
    • Google Scholar
96. 1. Salzberg SL

    (2017) Horizontal gene transfer is not a hallmark of the human genome

    Genome Biology 18:1–5.

    https://doi.org/10.1186/s13059-017-1214-2

    • PubMed
    • Google Scholar
97. 1. Šmarda P
    2. Bureš P
    3. Horová L
    4. Leitch IJ
    5. Mucina L
    6. Pacini E
    7. Tichý L
    8. Grulich V
    9. Rotreklová O

    (2014) Ecological and evolutionary significance of genomic GC content diversity in monocots

    PNAS 111:E4096–E4102.

    https://doi.org/10.1073/pnas.1321152111

    • PubMed
    • Google Scholar
98. Book

    1. Sokal R
    2. Rohlf FJ

    (1995)

    Biometry

    W. H Freeman.

    • Google Scholar
99. 1. Su Z
    2. Huang W
    3. Gu X

    (2011) Comment on "Positive selection of tyrosine loss in metazoan evolution"

    Science 332:917.

    https://doi.org/10.1126/science.1187374

    • PubMed
    • Google Scholar
100. 1. Tan CS
     2. Pasculescu A
     3. Lim WA
     4. Pawson T
     5. Bader GD
     6. Linding R

     (2009) Positive selection of tyrosine loss in metazoan evolution

     Science 325:1686–1688.

     https://doi.org/10.1126/science.1174301

     • PubMed
     • Google Scholar
101. 1. Tan CS
     2. Schoof EM
     3. Creixell P
     4. Pasculescu A
     5. Lim WA
     6. Pawson T

     (2011) Response to comment on ‘Positive Selection of Tyrosine Loss in Metazoan Evolution

     Science 332:917.

     https://doi.org/10.1126/science.1188535

     • Google Scholar
102. 1. Thornton JW
     2. DeSalle R

     (2000) Gene family evolution and homology: genomics meets phylogenetics

     Annual Review of Genomics and Human Genetics 1:41–73.

     https://doi.org/10.1146/annurev.genom.1.1.41

     • PubMed
     • Google Scholar
103. 1. Tien MZ
     2. Meyer AG
     3. Sydykova DK
     4. Spielman SJ
     5. Wilke CO

     (2013) Maximum allowed solvent accessibilities of residues in proteins

     PLOS ONE 8:e80635.

     https://doi.org/10.1371/journal.pone.0080635

     • PubMed
     • Google Scholar
104. 1. Toll-Riera M
     2. Bostick D
     3. Albà MM
     4. Plotkin JB

     (2012) Structure and age jointly influence rates of protein evolution

     PLOS Computational Biology 8:e1002542.

     https://doi.org/10.1371/journal.pcbi.1002542

     • PubMed
     • Google Scholar
105. 1. Tourasse NJ
     2. Li WH

     (2000) Selective constraints, amino acid composition, and the rate of protein evolution

     Molecular Biology and Evolution 17:656–664.

     https://doi.org/10.1093/oxfordjournals.molbev.a026344

     • PubMed
     • Google Scholar
106. 1. Trifonov EN

     (2000) Consensus temporal order of amino acids and evolution of the triplet code

     Gene 261:139–151.

     https://doi.org/10.1016/S0378-1119(00)00476-5

     • PubMed
     • Google Scholar
107. 1. Vakirlis N
     2. Hebert AS
     3. Opulente DA
     4. Achaz G
     5. Hittinger CT
     6. Fischer G
     7. Coon JJ
     8. Lafontaine I

     (2018) A molecular portrait of de novo genes in yeasts

     Molecular Biology and Evolution 35:631–645.

     https://doi.org/10.1093/molbev/msx315

     • PubMed
     • Google Scholar
108. 1. Vakirlis N
     2. Carvunis AR
     3. McLysaght A

     (2020) Synteny-based analyses indicate that sequence divergence is not the main source of orphan genes

     eLife 9:e53500.

     https://doi.org/10.7554/eLife.53500

     • PubMed
     • Google Scholar
109. 1. Van Oss SB
     2. Carvunis AR

     (2019) De novo gene birth

     PLOS Genetics 15:e1008160.

     https://doi.org/10.1371/journal.pgen.1008160

     • PubMed
     • Google Scholar
110. 1. Velander P
     2. Wu L
     3. Henderson F
     4. Zhang S
     5. Bevan DR
     6. Xu B

     (2017) Natural product-based amyloid inhibitors

     Biochemical Pharmacology 139:40–55.

     https://doi.org/10.1016/j.bcp.2017.04.004

     • PubMed
     • Google Scholar
111. Preprint

     1. Weibel C

     (2020) The protein domains of vertebrate species in which selection is more effective have greater intrinsic structural disorder

     bioRxiv.

     https://doi.org/10.1101/2020.10.15.341313

     • Google Scholar
112. 1. Weisman CM
     2. Murray AW
     3. Eddy SR

     (2020) Many, but not all, lineage-specific genes can be explained by homology detection failure

     PLOS Biology 18:e3000862.

     https://doi.org/10.1371/journal.pbio.3000862

     • PubMed
     • Google Scholar
113. 1. Weiss MC
     2. Sousa FL
     3. Mrnjavac N
     4. Neukirchen S
     5. Roettger M
     6. Nelson-Sathi S
     7. Martin WF

     (2016) The physiology and habitat of the last universal common ancestor

     Nature Microbiology 1:1–8.

     https://doi.org/10.1038/nmicrobiol.2016.116

     • Google Scholar
114. 1. Willis S
     2. Masel J

     (2018) Gene birth contributes to structural disorder encoded by overlapping genes

     Genetics 210:303–313.

     https://doi.org/10.1534/genetics.118.301249

     • PubMed
     • Google Scholar
115. 1. Wilson BA
     2. Foy SG
     3. Neme R
     4. Masel J

     (2017) Young genes are highly disordered as predicted by the preadaptation hypothesis of De Novo Gene Birth

     Nature Ecology & Evolution 1:0146.

     https://doi.org/10.1038/s41559-017-0146

     • PubMed
     • Google Scholar
116. 1. Wolfe K

     (2004) Evolutionary genomics: yeasts accelerate beyond BLAST

     Current Biology 14:R392–R394.

     https://doi.org/10.1016/j.cub.2004.05.015

     • PubMed
     • Google Scholar
117. 1. Zerbino DR
     2. Achuthan P
     3. Akanni W
     4. Amode MR
     5. Barrell D
     6. Bhai J
     7. Billis K
     8. Cummins C
     9. Gall A
     10. Girón CG
     11. Gil L
     12. Gordon L
     13. Haggerty L
     14. Haskell E
     15. Hourlier T
     16. Izuogu OG
     17. Janacek SH
     18. Juettemann T
     19. To JK
     20. Laird MR
     21. Lavidas I
     22. Liu Z
     23. Loveland JE
     24. Maurel T
     25. McLaren W
     26. Moore B
     27. Mudge J
     28. Murphy DN
     29. Newman V
     30. Nuhn M
     31. Ogeh D
     32. Ong CK
     33. Parker A
     34. Patricio M
     35. Riat HS
     36. Schuilenburg H
     37. Sheppard D
     38. Sparrow H
     39. Taylor K
     40. Thormann A
     41. Vullo A
     42. Walts B
     43. Zadissa A
     44. Frankish A
     45. Hunt SE
     46. Kostadima M
     47. Langridge N
     48. Martin FJ
     49. Muffato M
     50. Perry E
     51. Ruffier M
     52. Staines DM
     53. Trevanion SJ
     54. Aken BL
     55. Cunningham F
     56. Yates A
     57. Flicek P

     (2018) Ensembl 2018

     Nucleic Acids Research 46:D754–D761.

     https://doi.org/10.1093/nar/gkx1098

     • PubMed
     • Google Scholar
118. 1. Zile K
     2. Dessimoz C
     3. Wurm Y
     4. Masel J

     (2020) Only a single taxonomically restricted gene family in the Drosophila melanogaster subgroup can be identified with high confidence

     GBE Evaa 12:1355–1366.

     https://doi.org/10.1093/gbe/evaa127

     • Google Scholar
119. 1. Zuckerkandl E

     (1975) The appearance of new structures and functions in proteins during evolution

     Journal of Molecular Evolution 7:1–57.

     https://doi.org/10.1007/BF01732178

     • PubMed
     • Google Scholar

Author details

1. Jennifer E James

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing

   For correspondence

   jejames@email.arizona.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-0518-6783

2. Sara M Willis

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Data curation, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1605-6426

3. Paul G Nelson

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Data curation, Formal analysis, Methodology, Writing - original draft

   Competing interests

   No competing interests declared

4. Catherine Weibel

   1. Department of Physics, University of Arizona, Tucson, United States
   2. Department of Mathematics, University of Arizona, Tucson, United States

   Contribution

   Data curation, Software, Investigation, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-1837-5209

5. Luke J Kosinski

   Department of Molecular and Cellular Biology, University of Arizona, Tucson, United States

   Contribution

   Resources, Formal analysis, Methodology, Writing - review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-8146-5955

6. Joanna Masel

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Conceptualization, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing

   For correspondence

   masel@email.arizona.edu

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7398-2127

• 1,949

  views

• 212

  downloads

• 26

  citations

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