This is CoalHMM, a framework for demographic inference using a sequential Markov coalescence model. The framework is based on continuous time Markov chains of systems of two neighbouring nucleotides and allows for construction of complex demographic models, only limited by state space explosion (which actually is a rather hard limit when it comes to the number of populations and samples it can handle).

The framework has been used to construct two concrete models, an isolation model described in

T. Mailund, J. Y. Dutheil, A. Hobolth, G. Lunter, and M. H. Schierup, “Estimating
divergence time and ancestral effective population size of Bornean and Sumatran
orangutan subspecies using a coalescent hidden Markov model.,” PLoS Genet, vol. 7,
no. 3, p. e1001319, Mar. 2011.

and an isolation-with-migration model described in

T. Mailund, A. E. Halager, M. Westergaard, J. Y. Dutheil, K. Munch, L. N. Andersen,
G. Lunter, K. Prüfer, A. Scally, A. Hobolth, and M. H. Schierup, “A New Isolation
with Migration Model along Complete Genomes Infers Very Different Divergence
Processes among Closely Related Great Ape Species.,” PLoS Genet, vol. 8, no. 12,
pp. e1003125–e1003125, Nov. 2012.

These two models are available in scripts/estimate-using-I-model.py and scripts/estimate-using-IM-model.py, respectively.

Both models work on pairwise alignments. Given a pairwise alignment, it must first be translated into the compressed format used by zipHMM (http://ziphmm.googlecode.com) that we use for fast likelihood computations. The script prepare-alignments.py reads a number of different alignments formats and outputs an alignment in zipHMM format.

If you have full genome alignments you might want to split it into segments that you can handle in parallel, but the zipHMM should be fast enough to handle full chromosomes if you prefer.

Once you have formatted the alignment, you can run isolation or isolation-with-migration model using the estimate-using-I-model.py or estimate-using-IM-model.py scripts. Both scripts accept initial parameters for the likelihood maximisation and will output the estimated parameters.

See the script test_data/analysis.sh for an example of preparing alignments and running the two models.