pysinger

A pure-Python rewrite of the SINGER Bayesian ARG sampler.

Disclaimer

This project was set up mainly as an educational resource for myself to understand the SINGER algorithm by reimplementing it from scratch. All intellectual credit belongs to the original authors. For production use, the authoritative reference, and the optimised C++ implementation, please refer to the original repository:

github.com/popgenmethods/SINGER

Deng, Y., Nielsen, R. & Song, Y.S. Robust and accurate Bayesian inference of genome-wide genealogies for hundreds of genomes. Nature Genetics 57, 2124–2135 (2025). doi:10.1038/s41588-025-02317-9

• Overview
  • What pysinger does
  • The SINGER algorithm in one paragraph
  • Pipeline
  • Package map
• Build SINGER yourself
  • Step 1 — The Node class
  • Step 2 — The Branch class
  • Step 3 — The Tree class
  • Step 4 — The Recombination class
  • Step 5 — The Interval and IntervalInfo classes
  • Step 6 — The ARG class
  • Step 7 — The CoalescentCalculator
  • Step 8 — Emission models
  • Step 9 — The BSP (Branch Sequence Propagator)
  • Step 10 — The TSP (Time Sequence Propagator)
  • Step 11 — The Threader
  • Step 12 — The Sampler
  • Step 13 — VCF reader
  • Step 14 — Fitch parsimony reconstruction
  • Step 15 — tskit export
  • Putting it all together
• Mathematical background
  • 1. The coalescent with recombination
  • 2. The Branch Sequence Propagator (BSP)
  • 3. The Time Sequence Propagator (TSP)
  • 4. The MCMC
  • 5. Ancestral state reconstruction (Fitch parsimony)
  • 6. ARG encoding
• Data structures
  • Node
  • Branch
  • Tree
  • Recombination
  • Interval and IntervalInfo
  • ARG
• Hidden Markov Models
  • CoalescentCalculator
  • Emission models
  • BSP — Branch Sequence Propagator
  • TSP — Time Sequence Propagator
• MCMC threading
  • Threader
  • Sampler
• I/O and reconstruction
  • VCF reader
  • tskit writer
  • Rate maps
  • Fitch parsimony reconstruction
• API reference
  • Sampler
  • Data structures
  • HMM
  • MCMC
  • I/O
  • Rates
  • Reconstruction