sabifo4 / mammals_dating

Methods for the analyses described in Álvarez-Carretero, et al. 2022 (https://doi.org/10.1038/s41586-021-04341-1)

A Species-Level Timeline of Mammal Evolution Integrating Phylogenomic Data

What will you find here?

In this repository, you can find a description of the several steps we have carried out to infer the divergence times of a phylogeny with 4,705 mammal taxa using our new Bayesian sequential-subtree dating approach. In a nutshell, this approach consists of using the inferred posterior divergence times on a first dataset (in this study, this is a 72-taxon alignment generated with 15,268 filtered mammal genomes from ENSEMBL ordered from slow- to fast-evolving and partitioned into four blocks) as calibration priors for a subsequent dating analysis with a much larger dataset (in this study, this is a concatenated alignment of 168 nuclear, 12 mitochondrial protein-coding, and 2 mitochondrial non-coding genes for 4,705 mammal taxa). The advantage of this method is how much it can reduce the amount of computational time needed to date such a large phylogeny.

How is the content structured?

We have divided this repository into the following sections:

1. 00_Data_collection
   Every project starts with the data collection. You can follow the link above to read all the details about this step.
2. 01_SeqBayes_S1
   Once the data are gathered, they need to be processed and filtered before proceeding with Bayesian inference. Therefore, this directory contains several subdirectories in which all these steps are detailed:
   • 00_Gene_filtering
     Here, you will find all the analyses we carried out to filter the data and the steps we followed to then obtain the partitioned alignment with 72 taxa.
   • 01_BASEML
     Here, you will find the step-by-step tutorial we followed to calculate the Hessian and the gradient of the partitioned alignment.
   • 02_MCMCtree
     Here, you can find the results of the Bayesian dating approach using the approximate likelihood as implemented in MCMCtree. Besides, you will also find the analyses to evaluate chain convergence.
   • 03_Fit_ST_to_posteriors
     After averaging over the posterior time estimates sampled during the independent MCMCs, we can fit skew-t (ST) distributions to each of the nodes of the 72-taxon phylogeny. Here, you can find all the details about this approach.
3. 02_SeqBayes_S2
   Once we have the ST distributions fitted to each node, we can proceed to use them as prior calibrations when dating the larger mammal dataset with ~5,000 taxa (dataset 2, number of taxa prior to data curation). The tasks are also divided in different steps as it was done in the first step:
   • 00_Data_filtering
     This directory contains a summary of the procedure followed to (i) curate the alignments for each data subset when working with the second dataset and (ii) generate the corresponding phylogenies. For each data subset, a directory has been generated with a README.md file in which a step-by-step guideline details all the filtering steps followed.
   • 01_BASEML
     Once the alignments and trees are ready, we can compute the Hessian and the gradient for the subsequent Bayesian inference of divergence times. Click the link above to be redirected to the steps and results of this step.
   • 02_MCMCtree_updcrh
     Here, you will find the final results regarding the inferred posterior times for the mammal phylogeny with 4,705 taxa.
   • 03_Generate_final_mammal_tree
     Here, you will find the final mammal Tree of Life that we generated at the end of our Bayesian sequential-subtree dating approach.
4. 03_Extra_analyses
   This directory contains a description of extra analyses that were carried out after the main analyses were finished. The main purpose was to compare the divergence times inferred with the 72-taxon alignment (dataset 1) to those estimated when using (i) 182 loci, (ii) only nuclear (1st+2nd CPs) loci, and (iii) only mitochondrial (1st+ 2nd CPs) loci for the same 72 mammal taxa (genes extracted from dataset 2 for these 72 mammal taxa).
5. calibrations
   This directory contains a pdf file with the justification for the calibrations used.
6. figs
   This directory contains figures that are embedded in some of the README.md files in this GitHub repository.
7. src
   This directory contains the software needed to carry out the tasks described above.

Do you have a compiled version of your data?

Yes, we do! We have uploaded a zip file with the data that you need to reproduce our analyses in FigShare, which you can access at DOI 10.6084/m9.figshare.14885691.

In this zip file you will find several directories with the following content:

• aln: here you will find the alignments used in both steps of our Bayesian sequential-subtree approach. Directory aln/00_step_01/ contains the alignments for the 72 genomes in phylip format. Directory aln/01_step_02/ contains the uncompressed, raw subtree alignments for the 4,705 taxa, with missing species represented as sequences of gaps. Directory aln/01_step_02_patterns/ contains the same alignments after processing, that is, with missing species removed and with the alignments compressed into site patterns (see MCMCtree and PAML documentation for alignment formats). The processed alignments are the ones used by MCMCtree to calculate the likelihood during estimation of gradient and Hessian. Note each alignment file contains several alignment blocks, with each block corresponding to an alignment partition. If you load an alignment file into an alignment editor, make sure your editor allows you to see all partition blocks and not just the first one.
• clocktest: here, you will find the full results for the Bayesian selection of relaxed-clock model.
• inBV: here you will find the in.BV files generated by BASEML, which contain the estimated gradient and Hessian for each alignment partition, which are required to estimate the divergence times under the approximate likelihood method in MCMCtree. This directory is subdivided into step01 and step02 subdirectories corresponding to the data for the 72 genomes and the 4,705-taxon subtrees, respectively.
• paleodb: here you will find a csv file with the details about the mining of mammal genera from PaleoDB.
• timetrees: here you will find the timetrees estimated with MCMCtree in Nexus format. These are suitable for plotting with FigTree. For the 4,705 taxa, both the separate subtrees and the fully stitched 4,705-taxon tree are provided.
• trees: here you will find the calibrated tree topologies (input tree files needed by MCMCtree) in Newick format for each of the dataset (both from the first and the second step of our Bayesian approach). Subdirectories step01 and step02 contain the corresponding trees for the 72 genomes and 4,705 taxa, respectively. These trees are required, together with the alignment and in.BV files to estimate the divergence times with `MCMCtree.

In addition, please note that, in this GitHub repository, we have README.md files where we explain in detail each analysis we carried out during this study in a step-by-step tutorial manner. Apart from providing you with code snippets and links to the scripts and source code we used, we also include in these README.md files the links to zip files that you can use to reproduce each analysis (as well as to check if you have gotten the same results we did!).

---

We hope you can find this repository useful. Happy Bayesian inference! :)