Warning: This is rapidly evolving research software, and still in a largely experimental part of its life cycle. The methods should be reliable (see the tests), but the API may change. Contributions welcome, feel free to reach out.
DeadBird.jl is a julia package for modeling comparative genomic count data using phylogenetic birth-death processes, most commonly gene families. DeadBird.jl uses the (exact) algorithm of Csuros & Miklos (2009) for computing the conditional survival likelihoods.
Some things DeadBird.jl currently allows to do:
- Flexibly specify models of evolution (e.g. with branch-specific rates using molecular clock priors, models of rate heterogeneity across families, different prior distributions on the number of lineages at the root, ...).
- Perform Bayesian inference and maximum likelihood estimation (using automatic differentiation and the Turing library for probabilistic programming) of duplication, loss and gain rates for these complex models along a known phylogeny.
- Simulate data under these possibly complicated models, and assess model fit using posterior predictive simulations.
- Statistically test for whole-genome multiplications along branches of the species tree (sensu Rabier et al. 2014, Zwaenepoel & Van de Peer 2019).
DeadBird.jl is developed by Arthur Zwaenepoel (member of the Van de Peer group at VIB-UGent center for plant systems biology). If you use DeadBird.jl please cite the following article (which describes a previous version on which this package is based):
Zwaenepoel, A., and Y. Van de Peer.
"Model-based detection of whole-genome duplications in a phylogeny."
Molecular biology and evolution (2020).
and the article describing the likelihood algorithm for phylogenetic BDPs:
Csűrös, Miklós, and István Miklós.
"Streamlining and large ancestral genomes in Archaea inferred with a
phylogenetic birth-and-death model."
Molecular biology and evolution 26.9 (2009): 2087-2095.