RELAX: Detecting Relaxed Selection in a Phylogenetic Framework

In the paper

“Standard methods for estimating the strength and extent of purifying or positive selection from molecular sequence data are not suitable for detecting relaxed selection, because they lack power and can mistake an increase in the intensity of positive selection for relaxation of both purifying and positive selection.”

• Also stated in: § Introduction · plain

  “However, this test can be misleading because estimates of ! can be increased by processes such as intensification of positive selection when selective pressures vary both along branches and sites (Murrell et al. 2012).”

What the sources say

• Partly supported

  Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites Murrell et al. 2012 sourceExperimental — accuracy not yet measured

In the paper

“The most obvious approach has been to identify loss of trait or pseudogenization (e.g., Dainat et al. [2012]).”

What the sources say

• Supported

  The Abstract describes GLADX as an approach for detecting lineage-specific gene loss and pseudogenization, and says such losses could result from relaxed selection. The Introduction also describes gene-loss detection as a commonly used comparative method. This supports citing the paper as an example of the approach, though gene loss is not uniquely evidence of relaxed selection. (LLM, conf=0.88)

  cites Dainat et al. [2012] source

In the paper

“Relaxed selection has also been implicated as both a driver of speciation (Templeton 2008) and a precursor to extinction (Moran 1996).”

What the sources say

• ?Not checked — no full text of the sourcecites Templeton 2008 source
• Supported

  The citation to Moran (1996) supports the extinction part of the sentence, not the speciation part attributed to Templeton. The Discussion says that ineffective selection in small, asexual populations permits deleterious mutations to accumulate and reports that theoretical models link the resulting mutational load to possible extinction. It does not claim that Buchnera itself became extinct. (LLM, conf=0.85)

  cites Moran 1996 source

In the paper

“Such variation across the phylogeny can be studied using the BS-REL model (Kosakovsky Pond et al. 2011).”

What the sources say

• Supported

  The “Branch-Specific Branch-Site REL” section says every branch has its own ω values and mixing coefficients, estimated jointly from all sites. The “A New Test for Episodic Selection” section identifies this as a branch-site REL model, supporting the claim that it can study variation in the ω distribution across a phylogeny. (LLM, conf=0.94)

  cites Kosakovsky Pond et al. 2011 source

In the paper

“The introduction of branch-site models addressed evolutionary rate variation not only over sites but also over branches (Yang and Nielsen 2002).”

What the sources say

• Supported

  The Introduction says the authors implement models that allow the ω ratio to vary both among sites and among lineages. It contrasts these with earlier models that allowed variation in only one of those dimensions, supporting the claim about branch-site models. (LLM, conf=0.94)

  cites Yang and Nielsen 2002 source

In the paper

“Our random effects (Kosakovsky Pond et al. 2011) and mixed effects (Murrell et al. 2012) branch-site models allow detection of individual branches on which only a subset of sites are under positive selection or, conversely, individual sites which are under positive selection on only a subset of branches.”

What the sources say

• Partly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites Kosakovsky Pond et al. 2011 sourceExperimental — accuracy not yet measured
• Supported

  The Discussion states that MEME identifies selected sites by pooling information over branches, while the Introduction describes its branch-varying rates. This supports the claim attributed to Murrell et al.: detecting sites under positive selection on only a subset of branches. The separate claim about detecting selected branches is cited to another paper. (LLM, conf=0.90)

  cites Murrell et al. 2012 source

In the paper

“An !distribution can be thought of as an evolutionary fingerprint characterizing the evolutionary pattern of the gene, where a small subset of sites can have very large !values even when the majority of sites evolved under strong purifying selection (Kosakovsky Pond, Scheffler, et al. 2010).”

What the sources say

• ?Could not verify

  Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

  cites Kosakovsky Pond, Scheffler, et al. 2010 sourceExperimental — accuracy not yet measured

In the paper

“The introduction of site-to-site rate variation allowed detection of individual sites evolving under positive selection, by classifying sites into different rate categories and estimating a separate !value specifically for the sites in the positive selection category (Nielsen and Yang 1998).”

What the sources say

• Supported

  Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  cites Nielsen and Yang 1998 source

In the paper

“The first codon models estimated parameters representing an averaged nonsynonymous-synonymous rate ratio.”

What the sources say

• ?Not checked — no full text of the sourcecites Goldman and Yang 1994 source
• Partly supported

  The Probabilistic Model for Codon Evolution section says the model’s parameters determine synonymous and nonsynonymous substitution rates. But Parameter Interpretation describes the estimated parameters as products of instantaneous rates and time, not as an averaged nonsynonymous-to-synonymous rate ratio; its averaging over codons concerns expected substitution counts. (LLM, conf=0.82)

  cites Muse and Gaut 1994 sourceExperimental — accuracy not yet measured

In the paper

“Therefore, it became conventional to think in terms of the distribution of !values over sites instead of a single gene-wide !(Yang et al. 2000).”

What the sources say

• Partly supported

  The Theory section models site-to-site variation in the dN/dS ratio using a statistical distribution, and the Results compare those models with a one-ratio model. This supports the conceptual contrast in the claim, but the paper does not establish that thinking in distributions became conventional across the field. (LLM, conf=0.82)

  cites Yang et al. 2000 sourceExperimental — accuracy not yet measured

In the paper

“This development vastly increased the detectable instances of positive selection (Murrell et al. 2012).”

What the sources say

• Supported

  The Discussion reports that MEME detects additional sites beyond FEL, sometimes with dramatic differences, because it can identify positive selection on some branches where constant-selection models cannot. The paper supports the claimed increase in detectable instances, though its Results caution that pinpointing individual selected branches remains difficult. (LLM, conf=0.87)

  cites Murrell et al. 2012 source

In the paper

“Accordingly, traces of relaxed selection have been documented in bacterial endosymbionts, including fixation of deleterious alleles (e.g., loss of DNA replication proofreading enzymes) and overall elevated rates of nonsynonymous and radical amino acid replacements (Wernegreen 2011).”

What the sources say

• Partly supported

  The Introduction (page unavailable) reports genome-wide rate acceleration concentrated at nonsynonymous sites, and the Discussion reports elevated radical replacements and evidence consistent with fixation of slightly deleterious mutations. The copy does not document the cited example of lost DNA-replication proofreading enzymes. The Discussion also says relaxed selection may contribute but is unlikely to explain the genome-wide pattern on its own; it favors genetic drift and linkage. (LLM, conf=0.84)

  cites Wernegreen 2011 sourceExperimental — accuracy not yet measured

In the paper

“Insects can also carry facultative (secondary) endosymbionts whose fitness effects on their hosts can be variable and complex (Oliver et al. 2006(Oliver et al. , 2009;;Degnan et al. 2009;Burke et al. 2010).”

What the sources say

• Supported

  The paper’s opening summary (page not provided) identifies pea aphids as hosts of facultative, or secondary, symbionts. It reports both increased resistance to parasitism and severe fecundity costs from carrying two such symbionts, supporting the claim that their fitness effects can be variable and complex. (LLM, conf=0.93)

  cites Oliver et al. 2006 source
• Supported

  The paper describes H. defensa as an inherited aphid symbiont that is not required for survival or reproduction in the absence of parasitism (Abstract). It reports that protection varies substantially among strains and that infection can confer benefits under parasitoid pressure but carry costs without it, supporting the claim that fitness effects can be variable and complex. (LLM, conf=0.92)

  cites Oliver et al. 2009 source
• ?Not checked — no full text of the sourcecites Degnan et al. 2009 source

In the paper

“Many insect species have obligate (primary) bacterial endosymbionts that perform crucial metabolic processes for the host (Moran and Baumann 2000;McCutcheon et al. 2009;Vogel and Moran 2011).”

What the sources say

• ?Not checked — no full text of the sourcecites Moran and Baumann 2000 source
• Supported

  The introduction describes nutritional endosymbionts of insects as having mutually obligate relationships with their hosts. It also says symbionts supply compounds missing from insect diets and gives examples of amino-acid and vitamin biosynthesis in sap-feeding insects. (LLM, conf=0.91)

  cites McCutcheon et al. 2009 source
• ?Not checked — no full text of the sourcecites Vogel and Moran 2011 source

In the paper

“Borna disease viruses may have integrated into mammalian genomes over 65 Ma (Horie et al. 2013).”

What the sources say

• Partly supported

  The ‘Paleovirology of bornaviruses’ section says that bornavirus-like elements in several mammalian genomes appear to have been established approximately 65 Ma, and that bornavirus infection in mammals may go back to 65 Ma. It does not establish integration over 65 Ma ago. (LLM, conf=0.85)

  cites Horie et al. 2013 sourceExperimental — accuracy not yet measured

In the paper

“Recently, Tucker et al. (2013) examined the genomes of sexual and asexual D. pulex lineages and found evidence for more recent transitions to asexuality, on the order of decades.”

What the sources say

• Supported

  The paper reports sequencing 11 sexual and 11 asexual D. pulex genomes. Its lineage-age analysis estimates the average age of each asexual lineage at 22 years, supporting transitions on the order of decades; it distinguishes these young lineages from the older spread of the asexuality-associated haplotype. (LLM, conf=0.92)

  cites Tucker et al. (2013) source

In the paper

“however, asexual animal species are rare and typically do not persist for long periods of time (Barton 2009).”

What the sources say

• Partly supported

  The paper states that asexual species have a “relatively short life,” supporting the persistence part of the claim (section and page not provided in our copy). Its introduction says sexual reproduction dominates eukaryotic biology, but does not specifically establish that asexual animal species are rare. (LLM, conf=0.83)

  cites Barton 2009 sourceExperimental — accuracy not yet measured

In the paper

“RELAX, by avoiding the unrealistic single-!assumption, correctly identified intensified selection along test branches in 100/100 replicates, with a mean selection intensity parameter (k) of 3.79.”

In the paper

“Specifically, !T;1 , representing the strongest purifying selection, shifted up by an order of magnitude; !T;2 , representing weakly purifying selection shifted from 0.3 to 1.0; and !T;3 , representing positive selection, shifted down by an order of magnitude.”

In the paper

“Relative to free-living bacteria, there was sufficient signal to detect relaxation in both primary endosymbionts (k = 0.28; P-value < 0.0001) and secondary endosymbionts (k = 0.61; Pvalue < 0.0001) independently.”

In the paper

“The backbone of the phylogeny is under intense selection (k b 41 shown as red in fig.6), as are terminal branches leading to free-living -proteobacteria, whereas the endosymbiont lineages tend to be more relaxed (k b < 1 shown as blue/purple in fig.6).”

In the paper

“We demonstrated that RELAX is well behaved on simulated data and established its power to detect relaxed selection via simulation and in four biological systems predicted to exhibit changes in the intensity of selection on a subset of branches.”

In the paper

“We inferred 78% of the sites evolved under an ! of 0.82, with the remaining 22% evolving under !¼ 1:2.”

In the paper

“A higher !within a group is interpreted as evidence for relaxation of selective constraints, as those constraints tend to reduce the rate of nonsynonymous substitutions relative to that of synonymous substitutions, thereby keeping !values low.”

In the paper

“This approach detects the terminal effects of relaxation, but it is not capable of detecting subtler changes in the intensity of selection that may not drive a gene to obsolescence or cause a loss of function.”

In the paper

“We establish the validity of our test via simulations and show that it can distinguish between increased positive selection and a relaxation of selective strength.”

In the paper

“When rate variation over time is not modeled, !values for specific sites or site categories represent time-averages over all branches in the phylogeny.”

In the paper

“In the context of branch-site models with multiple !categories represented by different proportions of sites, this trend can have two different effects: The ! values inferred for the selection categories can move toward 1 (fig.1) and/or the proportions of sites belonging to the different categories can Detecting Relaxed Selection .doi:10.1093/molbev/msu400MBE change in such a way that more sites”

In the paper

“Such gene-wide average !values were typically below 1, because most sites in protein-coding genes are strongly conserved.”

In the paper

“This approach led to the detection of sites for which ! is well above 1, even in genes where the average !value estimated across all sites is below 1.”

In the paper

“When the introduction of k produces a statistically significant improvement in model fit, we conclude that selection on test branches is intensified (k 4 1) or relaxed (k < 1) compared with background branches.”

In the paper

“The change from phylogeny-level resolution to branch-level resolution yielded higher branch-specific !values than phylogeny-wide !values, mirroring the effect obtained when site-to-site rate variation was introduced.”

In the paper

“When selection is relaxed, the smaller !values increase toward 1, whereas !values above 1 decrease.”

In the paper

“We validate the performance of RELAX using simulations as well as four biological examples where relaxation would be expected according to evolutionary theory: 1) endosymbiosis in -proteobacteria, 2) color vision genes in bats, 3) viral endogenization in bornaviruses, and 4) asexual reproduction in crustaceans.”

In the paper

“If the null model is rejected by the LRT, it indicates that selection on the test branches is intensified (k 4 1) or relaxed (k < 1) compared with the reference branches.”

In the paper

“Importantly, these simulated scenarios capture only a tiny fraction of the breadth under which selection may actually operate.”

In the paper

“Three sets of simulations were performed to assess the validity of RELAX.”

In the paper

“In empirical data sets, we expect the signal for relaxed selection to be driven primarily by sites evolving under purifying rather than positive selection, which are by far the majority.”

In the paper

“We also investigated the robustness of RELAX to the effect of increasing only the selection intensity of positive selection in a subset of branches (i.e., model misspecification).”

In the paper

“We enforced this constraint due to the flat likelihood surface associated with high !values, which makes the larger empirical estimate for this parameter unreliable.”

In the paper

“Natural selection is less efficient in this setting, and Muller's Ratchet can lead to the accumulation of slightly deleterious mutations (Muller 1964).”

What the sources say

• ?Not checked — no full text of the sourcecites Muller 1964 source

In the paper

“These obligate endosymbionts have arisen multiple times among the -proteobacteria (Husn ık et al. 2011) and are passed vertically from mother to child, thereby introducing convergent features in the bacterial genomes: gene loss, smaller genomes, lower GC content, and reduced thermal stability of proteins (Wernegreen 2002).”

What the sources say

• ?Not matched

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

  cites Husn ık et al. 2011
• ?Not checked — no full text of the sourcecites Wernegreen 2002 source

In the paper

“Furthermore, the General Descriptive model should not be used to select partitions for formal RELAX hypothesis testing, because it would amount to using the data twice and substantially bias the test result.”

In the paper

“Phylogenetic inference that does not account for variable substitution rates arising from skewed GC content found in endosymbionts tends to force their lineages together in a phylogenetic tree, whereas nonhomogeneous models of molecular evolution suggest a larger number of independent origins of endosymbionts from free-living bacteria (Husn ık et al. 2011).”

What the sources say

• ?Not matched

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

  cites Husn ık et al. 2011

In the paper

“This pattern was expected since primary endosymbionts likely experience tighter bottlenecks during vertical transmission and their genomes show greater evidence of extreme evolutionary features (e.g., gene loss and AT bias; McCutcheon and Moran 2012).”

What the sources say

• ?Not checked — no full text of the sourcecites McCutcheon and Moran 2012 source

In the paper

“For all of the point estimates of k and ! reported below, it should be borne in mind that our statistical methodology allows us to determine the strength of evidence that k is different from 1 (reported as P-values) but not the sampling variances of the estimates.”

In the paper

“Zhao, Rossiter, et al. (2009) inferred lineage-specific !values and consistent patterns of purifying selection (! ranging between 0.0 and 0.48).”

What the sources say

• ?Not matched

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

  cites Zhao, Rossiter, et al. (2009)

In the paper

“The loss of vision in animals that reside in total darkness (e.g., cave-dwelling fish) is a classic case of trait loss caused by relaxed selection (Darwin 1859).”

What the sources say

• ?Not matched

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

  cites Darwin 1859

In the paper

“This relaxation of selection in all echolocating species, over tens of millions of years, has not led to the formation of M/LWS1 pseudogenes in the echolocating species, which suggests some functionality of this gene may have been retained.”

In the paper

“Therefore, one would expect an excess of mildly deleterious mutations accumulating in mitochondrial genomes due to relaxed selection in asexual lineages.”

In the paper

“This bias can be corrected by using nonhomogeneous nucleotide substitution models in phylogenetic inference (Galtier andGouy 1995, 1998).”

What the sources say

• ?Not matched

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

  cites Galtier andGouy 1995, 1998

In the paper

“A caveat of phylogenetic methods in general is that sequence data are uninformative regarding changes that happen at subbranch timescales and consequently that changes in selective pressure over short timescales are difficult to characterize.”

In the paper

“Mitochondria are not part of the sexually reproducing genome, but they should experience the selective effects of a transition from sexuality to asexuality (Paland and Lynch 2006).”

What the sources say

• ?Not checked — no full text of the sourcecites Paland and Lynch 2006 source

In the paper

“From an evolutionary perspective, sexual reproduction is costly (Maynard Smith 1978);”

What the sources say

• ?Not matched

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

  cites Maynard Smith 1978

In the paper

“These selective constraints likely arise from the necessity to encode genes for complex biological functions in compact genomes.”

In the paper

“The small deviation from neutrality may be driven by the components of these branches preceding viral endogenization.”