Welcome to sweepLink!

Joint inference of demography and linked selection from time-series genetic data.

SweepLink is a Bayesian framework for the joint inference of population sizes, migration rates, and linked selection directly from time-series allele count data (such as ancient DNA). By modelling the linkage between neighbouring loci, it detects selective sweeps with greater sensitivity and confidence than single-locus methods.

Why sweepLink?

Most existing tools estimate selection for each site in isolation, ignoring the broader genomic context and the hitchhiking effect, where selection on a causal variant also shifts the frequencies of nearby loci. When linkage is ignored, random genetic drift can mimic a sweep at individual loci, producing false positives, while weak sweeps go undetected. SweepLink addresses this by adding a genome-wide layer to its model to pool the evidence of selection across the genomic landscape.

• Joint inference – estimates demography (population size, migration) and locus-specific selection simultaneously.

• Full posterior distributions – as a Bayesian method, sweepLink returns the complete posterior distribution for every parameter rather than a single point estimate, providing a measure of uncertainty for each result.

• Linkage-aware – a two-layer HMM captures the correlation between adjacent loci, modelling the hitchhiking effect.

• High sensitivity to weak selection – recovers weak sweeps (\(s = 0.01\)) that single-locus tools miss.

• Rare false positives – under full joint inference of demography and selection, the false positive rate is around 0.1%, and such rare errors are predicted as weak selection.

• No arbitrary thresholds – sweepLink produces confident posterior probabilities, eliminating the need for empirically tuned discovery thresholds.

• Multi-population ready – the methodology and implementation support both single- and two-population scenarios.

Note

The performance figures quoted above (e.g. sensitivity to \(s = 0.01\) and a ~0.1% false positive rate) were obtained on human-like simulations (\(N=10{,}000\)). SweepLink itself is general and can be applied to any organism for which time-series allele count data are available.

Where to start

1. Installation — build it, and run the two-command quickstart to check the build works.

2. Tutorial — a complete analysis on a small bundled dataset, from a VCF to a list of loci under selection. Every number on those pages comes from actually running it.

3. Using your own data — the settings that stop being defaults and become your decisions.

The User Manual section then starts with how it works — what the model is actually doing, worth reading once — followed by every option of the executable and of the companion scripts, every file format, and a FAQ.

Citation

A manuscript describing sweepLink is in preparation. Citation details will be added here upon publication.