Recent advances in next-generation sequencing are allowing us to uncover the evolution of sex chromosomes in non-model organisms. This study [1] represents an example of this application to birds of two Sylvioidea species from the genus Zosterops (commonly known as white-eyes). The study is exemplary in the amount and types of data generated and in the thoroughness of the analysis applied. Both male and female genomes were sequenced to allow the authors to identify sex-chromosome specific scaffolds. These data were augmented by generating the transcriptome (RNA-seq) data set. The findings after the analysis of these extensive data are intriguing: neoZ and neoW chromosome scaffolds and their breakpoints were identified. Novel sex chromosome formation appears to be accompanied by translocation events. The timing of formation of novel sex chromosomes was identified using molecular dating and appears to be relatively recent. Yet first signatures of distinct evolutionary patterns of sex chromosomes vs. autosomes could be already identified. These include the accumulation of transposable elements and changes in GC content. The changes in GC content could be explained by biased gene conversion and altered recombination landscape of the neo sex chromosomes. The authors also study divergence and diversity of genes located on the neo sex chromosomes. Here their findings appear to be surprising and need further exploration. The neoW chromosome already shows unique patterns of divergence and diversity at protein-coding genes as compared with genes on either neoZ or autosomes. In contrast, the genes on the neoZ chromosome do not display divergence or diversity patterns different from those for autosomes. This last observation is puzzling and I believe should be explored in further studies. Overall, this study significantly advances our knowledge of the early stages of sex chromosome evolution in vertebrates, provides an example of how such a study could be conducted in other non-model organisms, and provides several avenues for future work.

References

[1] Leroy T., Anselmetti A., Tilak M.K., Bérard S., Csukonyi L., Gabrielli M., Scornavacca C., Milá B., Thébaud C. and Nabholz B. (2019). A bird’s white-eye view on neo-sex chromosome evolution. bioRxiv, 505610, ver. 4 peer-reviewed and recommended by PCI Evolutionary Biology. doi: 10.1101/505610

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