The preprint by Bertels et al. [1] reports an interesting application of the well-accepted idea that positively selected traits (here variants) can appear several times independently; think about the textbook examples of flight capacity. Hence, the authors assume that reciprocally convergence implies positive selection. The methodology becomes then, in principle, straightforward as one can simply count variants in independent datasets to detect convergent mutations.
In this preprint, the authors have applied this counting strategy on 95 available sequence alignments of the env gene of HIV-1 [2,3] that corresponds to samples taken in different patients during the early phase of infection, at the very beginning of the onset of the immune system. They have compared the number and nature of the convergent mutations to a "neutral" model that assumes (a) a uniform distribution of mutations and (b) a substitution matrix estimated from the data. They show that there is an excess of convergent mutations when compared to the “neutral” expectations, especially for mutations that have arisen in 4+ patients. They also show that the gp41 gene is enriched in these convergent mutations. The authors then discuss in length the potential artifacts that could have given rise to the observed pattern.
I think that this preprint is remarkable in the proposed methodology. Samples are taken in different individuals, whose viral populations were founded by a single particle. Thus, there is no need for phylogenetic reconstruction of ancestral states that is the typical first step of trait convergent analyses. It simply becomes counting variants. This simple counting procedure needs nonetheless to be compared to a “neutral” expectation (a reference model), which includes the mutational process. In this article, the poor predictions of a specifically designed reference model is interpreted as an evidence for positive selection.
Whether the few mutations that are convergent in 4-7 samples out of 95 were selected or not is hard to assess with certainty. The authors have provided good evidence that they are, but only experimental validations will strengthen the claim. Nonetheless, beyond a definitive clue to the implication of selection on these particular mutations, I found the methodological strategy and the discussions on the potential biases highly stimulating. This article is an excellent starting point for further methodological developments that could be then followed by large-scale analyses of convergence in many different organisms and case studies.

References

[1] Bertels, F., Metzner, K. J., & Regoes R. R. (2018). Convergent evolution as an indicator for selection during acute HIV-1 infection. BioRxiv, 168260, ver. 4 peer-reviewed and recommended by PCI Evol Biol. doi: 10.1101/168260
[2] Keele, B. F., Giorgi, E. E., Salazar-Gonzalez, J. F., Decker, J. M., Pham, K.T., Salazar, M. G., Sun, C., Grayson, T., Wang, S., Li, H. et al. (2008). Identification and characterization of transmitted and early founder virus envelopes in primary HIV-1 infection. Proc Natl Acad Sci USA 105: 7552–7557. doi: 10.1073/pnas.0802203105
[3] Li, H., Bar, K. J., Wang, S., Decker, J. M., Chen, Y., Sun, C., Salazar-Gonzalez, J.F., Salazar, M.G., Learn, G.H., Morgan, C. J. et al. (2010). High multiplicity infection by HIV-1 in men who have sex with men. PLoS Pathogens 6:e1000890. doi: 10.1371/journal.ppat.1000890

It is my pleasure to recommend the paper by Raynal et al. [1] about using random forest for parameter inference. There are two reviews about the paper, one review written by Dennis Prangle and another review written by myself. Both reviews were positive and included comments that have been addressed in the current version of the preprint.

The paper nicely shows that modern machine learning approaches are useful for Approximate Bayesian Computation (ABC) and more generally for simulation-driven parameter inference in ecology and evolution.

The authors propose to consider the random forest approach, proposed by Meinshausen [2] to perform quantile regression. The numerical implementation of ABC with random forest, available in the abcrf package, is based on the RANGER R package that provides a fast implementation of random forest for high-dimensional data.

According to my reading of the manuscript, there are 3 main advantages when using random forest (RF) for parameter inference with ABC. The first advantage is that RF can handle many summary statistics and that dimension reduction is not needed when using RF.

The second advantage is very nicely displayed in Figure 5, which shows the main result of the paper. If correct, 95% posterior credibility intervals (C.I.) should contain 95% of the parameter values used in simulations. Figure 5 shows that posterior C.I. obtained with rejection are too large compared to other methods. By contrast, C.I. obtained with regression methods have been shrunken. However, the shrinkage can be excessive for the smallest tolerance rates, with coverage values that can be equal to 85% instead of the expected 95% value. The attractive property of RF is that C.I. have been shrunken but the coverage is of 100% resulting in a conservative decision about parameter values.

The last advantage is that no hyperparameter should be chosen. It is a parameter free approach, which is desirable because of the potential difficulty of choosing an appropriate acceptance rate.

The main drawback of the proposed approach concerns joint parameter inference. There are many settings where the joint parameter distribution is of interest and the proposed RF approach cannot handle that. In population genetics for example, estimation of the severity and of the duration of the bottleneck should be estimated jointly because of identifiability issues. The challenge of performing joint parameter inference with RF might constitute a useful research perspective.
 

References
 

[1] Raynal L, Marin J-M, Pudlo P, Ribatet M, Robert CP, Estoup A. 2017. ABC random forests for Bayesian parameter inference. arXiv 1605.05537v4, https://arxiv.org/pdf/1605.05537
[2] Meinshausen N. 2006. Quantile regression forests. Journal of Machine Learning Research 7: 983-999. http://www.jmlr.org/papers/v7/meinshausen06a.html

Dating nodes in a phylogeny is an important problem in evolution and is typically performed by using molecular clocks and fossil age estimates [1]. The manuscript by Chauve et al. [2] reports a novel method, which uses lateral gene transfers to help ordering nodes in a species tree. The idea is that a lateral gene transfer can only occur between two species living at the same time, which indirectly informs on node relative ages in a phylogeny: the donor species cannot be more recent than the recipient species. Horizontal gene transfers are increasingly recognized as frequent, even in eukaryotes, and especially in micro-organisms that have little fossil records [3-7]. Yet, such an important source of information has been very rarely used so far for inferring relative node ages in phylogenies. In this context, the method by Chauve et al. [2] represents an innovative and original approach to a difficult problem. An obvious limitation of the approach is that it relies on inferences of horizontal transfers, which detection is in itself a difficult problem. Incomplete taxon sampling, or the extinction of the true donor lineage may render patterns difficult to interpret in a temporary fashion. Yet, for clades with no fossils this may be the only piece of information we have at hand, and the growing amount of sequence data is likely to minimize issues derived from incomplete sampling.

The developed method, MaxTiC (for Maximal Time Consistency) [2], represents a very nice application of theoretical developments on the well-known « Feedback Arc Set » computer science problem to the evolutionary question of ordering nodes in a phylogeny. MaxTiC uses as input a species tree and a set of time constraints based on lateral gene transfers inferred using other softwares, and minimizes conflicts between node ordering and these time constraints. The application of MaxTiC on simulated datasets indicated that node ordering was fairly accurate [2]. MaxTiC is implemented in a freely available software, which represents original and relevant contribution to the field of evolutionary biology.

References

[1] Donoghue P and Smith M, editors. 2003. Telling the evolutionary time. CRC press.

[2] Chauve C, Rafiey A, Davin AA, Scornavacca C, Veber P, Boussau B, Szöllősi GJ, Daubin V and Tannier E. 2017. MaxTiC: Fast ranking of a phylogenetic tree by Maximum Time Consistency with lateral gene transfers. bioRxiv 127548, ver. 6 of 6th November 2017. doi: 10.1101/127548

[3] Ropars J, Rodríguez de la Vega RC, Lopez-Villavicencio M, Gouzy J, Sallet E, Debuchy R, Dupont J, Branca A and Giraud T. 2015. Adaptive horizontal gene transfers between multiple cheese-associated fungi. Current Biology 19, 2562–2569. doi: 10.1016/j.cub.2015.08.025

[4] Novo M, Bigey F, Beyne E, Galeote V, Gavory F, Mallet S, Cambon B, Legras JL, Wincker P, Casaregola S and Dequin S. 2009. Eukaryote-to-eukaryote gene transfer events revealed by the genome sequence of the wine yeast Saccharomyces cerevisiae EC1118. Proceeding of the National Academy of Science USA, 106, 16333–16338. doi: 10.1073/pnas.0904673106

[5] Naranjo-Ortíz MA, Brock M, Brunke S, Hube B, Marcet-Houben M, Gabaldón T. 2016. Widespread inter- and intra-domain horizontal gene transfer of d-amino acid metabolism enzymes in Eukaryotes. Frontiers in Microbiology 7, 2001. doi: 10.3389/fmicb.2016.02001

[6] Alexander WG, Wisecaver JH, Rokas A, Hittinger CT. 2016. Horizontally acquired genes in early-diverging pathogenic fungi enable the use of host nucleosides and nucleotides. Proceeding of the National Academy of Science USA. 113, 4116–4121. doi: 10.1073/pnas.1517242113

[7] Marcet-Houben M, Gabaldón T. 2010. Acquisition of prokaryotic genes by fungal genomes. Trends in Genetics. 26, 5–8. doi: 10.1016/j.tig.2009.11.007

Nobody can ignore that a full understanding of evolution requires an integrated approach from both conceptual and methodological viewpoints. Although some life-history traits, e.g. flowering time, have long been receiving more attention than others, in many cases because the former are more workable than the latter, we must acknowledge that our comprehension about how evolution works is strongly biased and limited. In the Arabidopsis community, such an integration is making good progress as an increasing number of research groups worldwide are changing the way in which evolution is put to the test.

This manuscript [1] is a good example of that as the authors raise an important issue in evolutionary biology by combining gene expression and flowering time data from different sources. In particular, the authors explore how variation in flowering time, which determines lifespan, and host immunity defenses co-vary, which is interpreted in terms of co-evolution between the two traits. Interestingly, the authors go beyond that pattern by separating lifespan-dependent from lifespan–independent defense genes, and by showing that defense genes with variants known to impact fitness in the field are among the genes whose expression co-varies most strongly with flowering time. Finally, these results are supported by a simple mathematical model indicating that such a relationship can also be expected theoretically.

Overall, the readers will find many conceptual and methodological elements of interest in this manuscript. The idea that evolution is better understood under the scope of life history variation is really exciting and challenging, and in my opinion on the right track for disentangling the inherent complexities of evolutionary research. However, only when we face complexity, we also face its costs and burdens. In this particular case, the well-known co-variation between seed dormancy and flowering time is a missing piece, as well as the identification of (variation in) putative selective pressures accounting for the co-evolution between defense mechanisms and life history (seed dormancy vs. flowering time) along environmental gradients. More intellectual, technical and methodological challenges that with no doubt are totally worth it.

Reference

[1] Glander S, He F, Schmitz G, Witten A, Telschow A, de Meaux J. 2017. Assortment of flowering time and defense alleles in natural Arabidopsis thaliana populations suggests co-evolution between defense and vegetative lifespan strategies. bioRxiv ver.1 of June 19, 2017. doi: 10.1101/131136

How parasite virulence evolves is arguably the most important question in both the applied and fundamental study of host-parasite interactions. Typically, this research area has been progressing through the formalization of the problem via mathematical modelling. This is because the question is a complex one, as virulence is both affected and affects several aspects of the host-parasite interaction. Moreover, the evolution of virulence is a problem in which ecology (epidemiology) and evolution (changes in trait values through time) are tightly intertwined, generating what is now known as eco-evolutionary dynamics. Therefore, intuition is not sufficient to address how virulence may evolve.
In their classical model, Anderson and May [1] predict that the optimal virulence level results from a trade-off between increasing parasite load within hosts and promoting transmission between hosts. Although very useful and foundational, this model incurs into several simplifying assumptions. One of the most obvious is that it considers that hosts are infected by a single parasite strain/species. Some subsequent models have thus accounted for multiple infections, generally predicting that this will select for higher virulence, because it increases the strength of selection in the within-host compartment.
Usually, when attacked, hosts deploy defences to combat their parasites. In many systems, however, parasites can suppress the immune response of their hosts. This leads to prolonged infection, which is beneficial for the parasite. However, immunosuppressed hosts are also more prone to infection. Thus, multiple infections are more likely in a population of immunosuppressed hosts, leading to higher virulence, hence a shorter infection period. Thus, the consequences of immunosuppression for the evolution of virulence in a system allowing for multiple infections are not straightforward.
Kamiya et al.[2] embrace this challenge. They create an epidemiological model in which the probability of co-infection trades off with the rate of recovery from infection, via immunosuppression. They then use adaptive dynamics to study how either immunosuppression or virulence evolve in response to one another, to then establish what happens when they both coevolve. They find that when virulence only evolves, its evolutionary equilibrium increases as immunosuppression levels increase. In the reverse case, that is, when virulence is set to a fixed value, the evolutionarily stable immunosuppression varies non-linearly with virulence, with first a decrease, but then an increase at high levels of virulence. The initial decrease of immunosuppression may be due to (a) a decrease in infection duration and/or (b) a decrease in the proportion of double infections, caused by increased levels of virulence. However, as virulence increases, the probability of double infections decreases even in non-immunosuppressed hosts, hence increased immunosuppression is selected for.
The combination of both Evolutionary Stable Strategies (ESSs) yields intermediate levels of virulence and immunosuppression. The authors then address how this co-ESS varies with host mortality and with the shape of the trade-off between the probability of co-infection and the rate of recovery. They find that immunosuppression always decreases with increased host mortality, as it becomes not profitable to invest on this trait. In contrast, virulence peaks at intermediate values of host mortality, unlike the monotonical decrease that is found in absence of immunosuppression. Also, this relationship is predicted to vary with the shape of the trade-off underlying the costs and benefits of immunosuppression.
In sum, Kamiya et al. [2] provide a comprehensive analysis of an important problem in the evolution of host-parasite interactions. The model provides clear predictions, and thus can now be tested using the many systems in which immunosuppression has been detected, provided that the traits that compose the model can be measured.

References

[1] Anderson RM and May RM. 1982. Coevolution of hosts and parasites. Parasitology, 1982. 85: 411–426. doi: 10.1017/S0031182000055360

[2] Kamiya T, Mideo N and Alizon S. 2017. Coevolution of virulence and immunosuppression in multiple infections. bioRxiv, ver. 7 peer-reviewed by PCI Evol Biol, 149211. doi: 10.1101/139147