RERconverge Expansion: Using Relative Evolutionary Rates to Study Complex Categorical Trait Evolution

RERconverge uses permulations, which generate null phenotype values using simulations restricted by the observed phenotype distribution to overcome the limitation of simulations.

“RERconverge uses permulations, which generate null phenotype values using simulations restricted by the observed phenotype distribution to overcome that limitation of simulations.”

• cites Saputra et al. 2021 sourceSupported

There is no corresponding group of internal branches with negative RERs assigned as omnivores-or any other category-in the categorical reconstruction.

“There is no corresponding group of internal branches with negative RERs assigned as omnivores—or any other category—in the categorical reconstruction”

The distributions of permulation P-values for the carnivore/omnivore and omnivore/carnivore binary analyses differed the most from that of the corresponding categorical post hoc test.

“We focused on the carnivore/omnivore and omnivore/carnivore binary analyses because their distributions of permulation P-values differed the most from that of the corresponding categorical post hoc test (supplementary fig. S4b, Supplementary Material online)”

The extension to RERconverge infers the ancestral evolutionary history of categorical traits.

“The extension to RERconverge presented here provides three updates enabling the use of the software for analyses of categorical traits: (i) it infers the ancestral evolutionary history of categorical traits”

Maximum likelihood ancestral trait reconstruction using a continuous time Markov model of evolution was implemented for categorical traits.

“we implemented maximum likelihood ancestral trait reconstruction using a continuous time Markov model of evolution; a framework that has been commonly applied to categorical traits (Pagel 1994; Pupko et al. 2000; Paradis et al. 2004).”

• cites Pagel 1994 sourceNot checked — no full text of the source
• cites Pupko et al. 2000 sourceNot checked — no full text of the source
• cites Paradis et al. 2004 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

Digestive tract development was significantly enriched in the carnivore/herbivore categorical post hoc test (permulation P = 0.0011).

“Digestive tract development was significantly enriched in the carnivore/herbivore categorical post hoc test (permulation P = 0.0011)”

The signal for PHB2 in the binary analysis is driven by a group of internal branches assigned as omnivores with negative RERs.

“The signal for this gene in the binary analysis is being driven by a group of internal branches assigned as omnivores with negative RERs”

The extension to RERconverge has options for both parametric and nonparametric statistical tests to associate RERs with categorical phenotypes, including post hoc pairwise testing.

“(iii) it has options for both parametric and nonparametric statistical tests to associate RERs with categorical phenotypes, including post hoc pairwise testing.”

Coercion to a binary trait has additional consequences on the calculation of relative evolutionary rates if species are removed.

“Coercion to a binary trait has additional consequences on the calculation of relative evolutionary rates if species are removed.”

When the group of internal branches with negative RERs is no longer present, there is no significant difference in RERs between carnivores and omnivores.

“When this group of internal branches with negative RERs is no longer present, there is no significant difference in RERs between carnivores and omnivores”

Under the ARD model, the ancestral mammal was inferred to be a carnivore.

“Under the ARD model, the ancestral mammal was inferred to be a carnivore”

In the categorical method, the RERs of carnivores are driven upwards due to the presence of internal branches with large RERs assigned as carnivores.

“Separating the relative evolutionary rates by terminal versus internal branches, we see that in the categorical method, the RERs of carnivores are driven upwards due to the presence of internal branches with large RERs assigned as carnivores”

MADCAM1 is significantly accelerated in omnivores according to the carnivore/omnivore binary analysis (permulation P = 0.0067).

“MADCAM1 is significantly accelerated in omnivores according to the carnivore/omnivore binary analysis (permulation P = 0.0067)”

PHB2 is not significantly accelerated in carnivores according to the categorical post hoc test (permulation P = 0.0776).

“but not in the categorical post hoc test (permulation P = 0.0776).”

Effects of the ancestral reconstruction extend beyond the gene level results to the pathway enrichment results.

“Effects of the ancestral reconstruction extend beyond the gene level results to the pathway enrichment results as well.”

MADCAM1 is not significantly accelerated in omnivores according to the corresponding categorical post hoc test (permulation P = 0.2161).

“however this is not the case in the corresponding categorical post hoc test (permulation P = 0.2161).”

PHB2 is significantly accelerated in carnivores according to the omnivore/carnivore binary analysis (permulation P < 1 × 10−4).

“PHB2 is significantly accelerated in carnivores according to the omnivore/carnivore binary analysis (permulation P < 1 × 10−4)”

Applying the new methodology to mammalian diet phenotypes, multiple diet-related pathways were enriched for genes with significant differences in RER distributions across diet categories.

“Applying our new methodology, we found multiple diet-related pathways enriched for genes with significant differences in RER distributions across diet categories”

Under the ARD model, omnivores and herbivores were inferred to have evolved independently multiple times.

“while omnivores and herbivores were inferred to have evolved independently multiple times”

In both selected genes, whether a rate shift is observed or not is driven by differences in the RER distributions of the internal branches.

“In both genes, whether a rate shift is observed or not is driven by differences in the RER distributions of the internal branches”

The extension to RERconverge provides a permulation strategy for categorical traits to return reliable, phylogeny aware, corrected P-values.

“(ii) it provides a permulation strategy for categorical traits to return reliable, phylogeny aware, corrected P-values”

The choice to remove a category of species from the analysis or to assign it to the background has a different impact on the ancestral inference.

“Which strategy is used to coerce the multi-categorical diet trait into a binary trait has different consequences on the default binary ancestral reconstruction.”

Genes ranked lowest according to the delta statistic but exhibiting a rate shift tended to be those evolving faster in carnivores compared to herbivores or omnivores.

“Genes in the pathway which were ranked lowest according to the delta statistic, but which exhibited a rate shift, tended to be those evolving faster in carnivores compared to herbivores or omnivores”

PIEZO1 has a significant difference in evolutionary rates across diet groups as detected by categorical RERconverge (omnibus test permulation P = 0.029).

“significant difference in evolutionary rates across groups as detected by our categorical RERconverge method (omnibus test permulation P = 0.029)”

The paper created a ranked list of genes that have convergently evolved with diet across mammals.

“we created a ranked list of genes that have convergently evolved with diet across mammals (supplementary data S2, Supplementary Material online).”

• cites supplementary data S2Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

phylANOVA has much lower power to detect genes with convergent shifts in evolutionary rates than categorical RERconverge when the same number (10,000) simulations/permulations are performed.

“phylANOVA has much lower power to detect genes with convergent shifts in evolutionary rates than categorical RERconverge when the same number (10,000) simulations/permulations are performed.”

These genes were enriched for pathways like "abnormal digestion" and "digestive tract development".

“These genes were enriched for pathways like “abnormal digestion” and “digestive tract development””

• cites supplementary data S3Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Other integrin-related genes were among the most significant genes identified by the categorical RERconverge analysis.

“other integrin-related genes were among the most significant genes identified by the categorical RERconverge analysis”

The number of categories in the analysis had a drastic effect on permulation time.

“the number of categories in the analysis had a drastic effect”

After permulations, starch and sucrose metabolism was retained as the top result (permulation P = 0.0021).

“After permulations, starch and sucrose metabolism was retained as the top result (permulation P = 0.0021)”

The genes driving the signal differed between the delta statistic and categorical RERconverge.

“but genes driving the signal differed between the two methods”

Analysis of the PIEZO1 alignment identified at least one position with site-level differences in conservation between diet categories.

“An analysis of the PIEZO1 alignment identified at least one position with site-level differences in conservation between diet categories.”

The ANOVA approach performed poorly in finding diet signal relative to the RERconverge updates and the delta statistic.

“it performed poorly in finding diet signal relative to our RERconverge updates and the delta statistic”

phylANOVA vastly underperforms compared to categorical RERconverge.

“we show that phylANOVA vastly underperforms compared to our categorical RERconverge.”

In the case of the diet phenotype, the delta statistic was biased toward capturing signal driven by slower evolution of carnivore lineages compared to other lineages.

“In the case of the diet phenotype, the delta statistic was biased toward capturing signal driven by slower evolution of carnivore lineages compared to other lineages.”

Significant differences in evolutionary rates for genes in the abnormal digestion pathway arise among all three pairwise comparisons, not just one.

“Second, significant differences in evolutionary rates for genes in this pathway arise among all three pairwise comparisons, not just one.”

The runtime increase grows more pronounced as additional categories are added even while keeping the set of species the same.

“This analysis provided further evidence that the runtime increase grows more pronounced as additional categories are added even while keeping the set of species the same”

Olfactory signaling and olfactory transduction were two of the strongest pathway enrichment results according to non-permulated P-values (P = 3.49 × 10−5 and 6.94 × 10−5).

“olfactory signaling and olfactory transduction were two of the strongest pathway enrichment results according to non-permulated P-values (P = 3.49 × 10−5 and 6.94 × 10−5).”

phylANOVA was less effective compared to binary and categorical RERconverge at detecting diet-relevant enriched pathways.

“we demonstrated that phylANOVA was less effective compared to binary and categorical RERconverge at detecting diet-relevant enriched pathways”

ITGB1 is evolving slower in carnivores compared to herbivores and omnivores.

“ITGB1 is evolving slower in carnivores compared to herbivores and omnivores”

The genes in the abnormal digestion pathway with the largest effect sizes identified by the delta statistic did not correspond with the most significant genes in this pathway identified by omnibus categorical RERconverge.

“The genes in the abnormal digestion pathway with the largest effect sizes identified by the delta statistic did not correspond with the most significant genes in this pathway identified by omnibus categorical RERconverge (Fig. 6a).”

The PIEZO1 residue corresponding to human residue 1,835 is more variable among herbivores and omnivores.

“this residue is more variable among herbivores and omnivores”

The current implementation of categorical permulations has limitations: the rejection sampler becomes very slow for large trees with greater than three categories, especially in phenotypes more highly clustered according to phylogeny.

“the current implementation of this method has limitations. The rejection sampler (see Methods, step (i)) becomes very slow for large trees with greater than three categories, especially in phenotypes which are more highly clustered according to phylogeny.”

The largest delta statistic genes tended to be those evolving slower in carnivores compared to omnivores or herbivores.

“The largest delta statistic genes tended to be those which were evolving slower in carnivores compared to omnivores or herbivores”

The categorical omnibus test can identify an enriched pathway of relevance that is not identified by any one binary or categorical pairwise/post hoc comparison, regardless of the method of ancestral reconstruction.

“We demonstrate that the categorical omnibus test can identify an enriched pathway of relevance that is not identified by any one binary or categorical pairwise/post hoc comparison, regardless of the method of ancestral reconstruction.”

For the abnormal digestion pathway, the enrichment results were similar between the categorical omnibus test and the delta statistic approach.

“for the “abnormal digestion” pathway, the enrichment results were similar between the categorical omnibus test we developed and the delta statistic approach.”

Using a small relaxation of the phenotype matching criterion (10%) can significantly speed up permulations, especially as category numbers increase.

“using a small relaxation of the phenotype matching criterion (10%) can significantly speed up the permulations, especially as category numbers increase”

There is no enrichment of the top delta statistic genes among the omnibus results.

“There is no enrichment of the top delta statistic genes among the omnibus results”

Using the CTMM ancestral reconstruction caused the enrichment of digestive tract development genes to increase in all binary tests except the herbivore/carnivore binary analysis.

“Using the new ancestral reconstruction caused the enrichment of digestive tract development genes to increase in all of the binary tests except for the herbivore/carnivore binary analysis such that their fold enrichment curves more closely matched that of the categorical post hoc test (Fig. 4a and b).”

Genes with greater phylogenetic signal for the diet phenotype tend to evolve slower in carnivores compared to herbivores.

“Genes with greater phylogenetic signal for the diet phenotype tend to evolve slower in carnivores compared to herbivores”

After permulations, olfactory signaling and olfactory transduction were no longer among the top enriched pathways (permulation P = 0.2866 and 0.3320).

“while olfactory signaling and olfactory transduction were no longer among the top enriched pathways (permulation P = 0.2866 and 0.3320)”

Categorical permulations have the same ability to refine pathway enrichment results.

“We also demonstrate that categorical permulations have the same ability to refine pathway enrichment results.”

The new categorical method outperforms phylogenetic simulations at identifying genes and enriched pathways significantly associated with the diet phenotypes.

“Our results show that our new categorical method outperforms phylogenetic simulations at identifying genes and enriched pathways significantly associated with the diet phenotypes”

The lack of enrichment for the pairwise binary analyses of the abnormal digestion pathway remained even when performed with the CTMM ancestral reconstruction.

“The lack of enrichment for the pairwise binary analyses remained even when performed with the CTMM ancestral reconstruction as was done with the digestive tract development pathway”

For analyses with three or more categories, species number appeared to have little to no effect on permulation runtime (R2 = 0.03, 0.14, 0.01, respectively).

“For analyses with three or more categories, species number appeared to have little to no effect on permulation runtime (R2 = 0.03, 0.14, 0.01, respectively)”

Olfactory signaling and olfactory transduction had lower non-permulated P-values than the starch and sucrose metabolism pathway from the canonical gene set (P = 1.00 × 10−4).

“These had lower P-values than the starch and sucrose metabolism pathway from the canonical gene set (P = 1.00 × 10−4)”

Multiple top genes within the abnormal digestion pathway have individual signal driven by differences in evolutionary rates between carnivores/omnivores and/or herbivores/omnivores.

“there are multiple top genes within the abnormal digestion pathway whose individual signal is being driven by differences in evolutionary rates between carnivores/omnivores and/or herbivores/omnivores instead”

TREH is evolving faster in carnivores compared to herbivores and omnivores.

“TREH is evolving faster in carnivores compared to herbivores and omnivores”

In those binary tests with increased fold enrichment, the Benjamini-Hochberg corrected P-values decreased.

“In those tests with increased fold enrichment, the Benjamini–Hochberg corrected P-values decreased as well.”

Using the CTMM for ancestral reconstruction resulted in increased enrichment for a pathway with clear relevance to the diet phenotype.

“Using the CTMM for ancestral reconstruction resulted in increased enrichment for a pathway with clear relevance to the diet phenotype”

Olfactory genes commonly cluster together in rank.

“we observed that olfactory genes commonly cluster together in rank”

The residue corresponding to residue 1,835 in the human PIEZO1 amino acid sequence is glycine in all but two of the 29 carnivore lineages, ignoring lineages with a gap character.

“The residue corresponding to residue 1,835 in the human PIEZO1 amino acid sequence is glycine in all but two of the 29 carnivore lineages (ignoring lineages with a gap character).”

The delta statistic lacks a method for computing P-values that takes phylogenetic relationships into account.

“and it lacks a method for computing P-values that takes phylogenetic relationships into account.”

The mammalian diet type phenotype has three main categories: herbivore, omnivore, and carnivore.

“the mammalian diet type phenotype has three main categories (herbivore, omnivore, and carnivore) that cannot be meaningfully ordered.”

• Also stated in: § Introduction · hedged

  “One common way of categorizing mammalian diets is by trophic level: carnivory, herbivory, and omnivory (Kim et al. 2016).”

• cites Kim et al. 2016 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

To study a non-ordinal categorical phenotype using most existing methods would require coercing the phenotype into binary groupings.

“To study such a phenotype using most existing methods would require coercing the phenotype into binary groupings, which is not straightforward”

• AssumptionRests on an assumption: The three mammalian diet categories (herbivore, omnivore, carnivore) cannot be meaningfully ordered.

Many comparative genomics methods lack the ability to analyze non-ordinal categorical traits with more than two categories.

“Many of these methods lack the ability to analyze non-ordinal categorical traits with more than two categories.”

• Also stated in: § Introduction · plain

  “However, a limitation to most phylogenetic comparative methods is their inability to analyze non-ordinal, categorical traits with more than two categories.”

Using a likelihood ratio test, the all rates different (ARD) model was selected to infer the ancestral states.

“Using a likelihood ratio test, as implemented in the expanded categorical RERconverge (Pagel 1994; Kowalczyk et al. 2019), we selected the all rates different (ARD) model to infer the ancestral states”

• cites Pagel 1994 sourceNot checked — no full text of the source
• cites Kowalczyk et al. 2019 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

Data in a phylogenetic context always violate the assumption of traditional statistical testing that each observation is independent because the data are related based on shared evolutionary history.

“Data in a phylogenetic context always violate the assumption of traditional statistical testing that each observation is independent because the data are related to each other based on their shared evolutionary history represented by their phylogeny.”

The delta statistic cannot be applied in the specific context of finding genetic elements that are convergently accelerated or conserved because it does not provide information on evolutionary rates.

“However, the delta statistic cannot be applied in the specific context of finding genetic elements that are convergently accelerated or conserved as it does not provide information on evolutionary rates”

Some of the internal branches with large RERs assigned as carnivores were likely assigned as omnivores or excluded in the binary analysis.

“some of these branches were likely assigned as omnivores or excluded in the binary analysis”

Differences in internal branch assignment could be driving the difference in apparent signal between the binary and categorical methods.

“suggesting that something such as differences in internal branch assignment could be driving the difference in the apparent signal of these methods.”

RERconverge can infer the ancestral history of a trait to use when associating evolutionary rates with the evolution of the trait.

“RERconverge can also infer the ancestral history of a trait to use when associating evolutionary rates with the evolution of the trait.”

Comparing distributions of RERs across phenotypes enables identification of regions of the genome showing convergent rate shifts in association with the phenotype.

“Comparing the distributions of these RERs across phenotypes enables us to identify regions of the genome showing convergent rate shifts in association with the phenotype”

These results suggest that the categorical method may help to control for false positives.

“Though the true ancestral phenotypes and evolutionary rates may remain unknown, these results suggest that the categorical method may help to control for false positives.”

The upward drive of carnivore RERs in the categorical method leads to no significant difference between the distributions of RERs in carnivores and omnivores in the categorical post hoc test.

“This leads to no significant difference between the distributions of RERs in carnivores and omnivores in the categorical post hoc test.”

The absence of the group of internal branches with negative RERs in the categorical reconstruction is likely because removing all herbivores affects the computation of RERs.

“likely because removing all herbivores affects the computation of RERs.”

Categorical RERconverge allows comparison of relative evolutionary rates of multiple phenotype categories simultaneously.

“the addition of categorical statistical tests in categorical RERconverge allows for the comparison of RERs of multiple phenotype categories simultaneously.”

The delta statistic does not use evolutionary rates.

“the delta statistic does not use evolutionary rates”

• cites Borges et al. 2019 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

• cites Ribeiro et al. 2023 sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

The enrichment of these genes for pathways like "abnormal digestion" and "digestive tract development" suggests that there is substantial signal within the analysis.

“which suggest that there is substantial signal within the analysis (supplementary data S3, Supplementary Material online).”

• cites supplementary data S3Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The results suggest that the delta statistic may be biased toward detecting genes with patterns of evolution driven by single pairwise comparisons evolving in one direction.

“Although the results are only explored for one trait, our results would suggest that the delta statistic may be biased toward detecting genes with patterns of evolution driven by single pairwise comparisons evolving in one direction.”

The categorical omnibus test is capturing rate shifts spread out across multiple pairwise comparisons which alone are not sufficient to detect enrichment for the abnormal digestion pathway.

“This suggests that the categorical omnibus test is capturing rate shifts spread out across multiple pairwise comparisons which alone aren’t sufficient to detect enrichment for this pathway.”

When enrichment is driven by a single pairwise comparison, an omnibus test is not necessarily expected to improve the strength of enrichment.

“In the case of enrichment being driven by a single pairwise comparison, as with digestive tract development, we would not necessarily expect an omnibus test to improve the strength of the enrichment.”

Whether more than one pairwise comparison contributes to enrichment may depend on which genes are experiencing changes in selective pressure in each phenotype state and on the statistical power of each pairwise test.

“Whether more than one pairwise comparison is contributing to enrichment may depend on which genes are experiencing changes in selective pressure in each phenotype state and on the statistical power of each pairwise test.”

When multiple pairwise comparisons drive enrichment, using a categorical omnibus test improves power to detect the pathway.

“when multiple pairwise comparisons are driving the enrichment, we expect that using a categorical omnibus test will improve our power to detect the pathway, and indeed this was the case with the abnormal digestion pathway.”

Results on the abnormal digestion pathway demonstrate that categorical analysis can identify convergent evolution in relevant genes even when a series of binary pairwise measurements fail.

“Our results on the “abnormal digestion” pathway demonstrate that categorical analysis can identify convergent evolution in relevant genes even when a series of binary pairwise measurements fail”

In mammalian diets, slower evolutionary rates among carnivores may improve the certainty of inferring carnivorous ancestors.

“In the case of mammalian diets, slower evolutionary rates among carnivores may improve the certainty of inferring carnivorous ancestors, corresponding to what we would expect from our maximum likelihood reconstruction on the master tree (supplementary fig. S1, Supplementary Material online), and leading to higher delta statistics in genes with slower evolutionary rates in carnivores.”

A 10% relaxation is not expected to affect the plausibility of the phenotype structures of the permulated trees.

“and is not expected to affect the plausibility of the phenotype structures of the permulated trees”

The omnibus test is one-sided, so high ranking genes within the same pathway contribute to gene set enrichment regardless of the direction of the rate shift for each gene.

“The omnibus test is one-sided, so high ranking genes within the same pathway contribute to gene set enrichment, regardless of the direction of the rate shift for each gene.”

Important information would be lost by excluding the carnivore/omnivore and herbivore/omnivore comparisons.

“emphasizes that important information would be lost by excluding them.”

Categorical RERconverge can detect rate shifts when either all or a subset of categories have those shifts.

“This allows the method to detect rate shifts when either all or a subset of categories have those shifts”

Carnivores may be less tolerant to changes at the PIEZO1 residue corresponding to human residue 1,835 compared to herbivores and omnivores.

“This suggests that carnivores may be less tolerant to changes in this position compared to herbivores and omnivores.”

Binary RERconverge determines enrichment using a Wilcoxon rank sum test to identify whether genes within a pathway are together shifted toward accelerated or slower evolutionary rates in foreground lineages.

“Binary RERconverge determines enrichment using a Wilcoxon rank sum test to identify whether genes within a pathway are together shifted toward accelerated evolutionary rates or toward slower evolutionary rates in foreground lineages.”

The paper extends the RERconverge method to include the analysis of categorical traits.

“Here, we extend the RERconverge method to include the analysis of categorical traits.”

• AssumptionRests on an assumption: There is a lack of available comparative tools to study categorical traits.

Results on the digestive tract development pathway demonstrate the importance of improving ancestral trait reconstruction with CTMM, even when binary phenotypes are used.

“Our results on the “digestive tract development” pathway demonstrate the importance of improving ancestral trait reconstruction with CTMM, even when binary phenotypes are used”

Improved certainty of inferring carnivorous ancestors may lead to higher delta statistics in genes with slower evolutionary rates in carnivores.

“In the case of mammalian diets, slower evolutionary rates among carnivores may improve the certainty of inferring carnivorous ancestors, corresponding to what we would expect from our maximum likelihood reconstruction on the master tree (supplementary fig. S1, Supplementary Material online), and leading to higher delta statistics in genes with slower evolutionary rates in carnivores.”

The differences in enrichment between binary tests were driven almost entirely by the ancestral reconstruction.

“Thus, we infer the differences in enrichment were being driven almost entirely by the ancestral reconstruction.”