Timepieces
Every great watch begins as a pile of parts on the bench.
Each Timepiece is a complete algorithm from population genetics, disassembled and rebuilt from scratch. Like a watchmaker laying out parts on the bench, we examine every gear before assembling the whole mechanism. Nothing is taken on faith. Nothing is hidden behind “see supplementary materials.”
Verification Status
Disclaimer
The code examples in each Timepiece are verified by automated unit tests that re-implement the documented functions and check their mathematical properties. No Timepiece has been independently verified by a domain expert. If you find an error – mathematical, pedagogical, or computational – please open an issue. The table below shows the current verification status.
The Timepieces are grouped by what they do, with verification status for each.
Simulators – tools that generate ground truth
# |
Timepiece |
Tests |
Verified |
What it does |
|---|---|---|---|---|
IV |
190 |
– |
Neutral coalescent with recombination. The clockwork that generates ground truth. |
|
XVI |
67 |
– |
Forward-time simulation with natural selection. The forge that builds what the coalescent cannot. |
|
XVIII |
132 |
– |
Coalescent simulation with selective sweeps via trajectory + structured coalescent. |
Demographic inference – estimating population size history
# |
Timepiece |
Tests |
Verified |
What it does |
|---|---|---|---|---|
I |
186 |
– |
Population size history from a single diploid genome. The simplest inference Timepiece. |
|
II |
112 |
– |
Extends PSMC to multiple unphased genomes with a distinguished lineage approach. |
|
XIII |
107 |
– |
Ultrafast pairwise TMRCA inference with gamma-distributed posteriors. |
|
XIV |
130 |
– |
GPU-accelerated Bayesian inference of population size history via SVGD. |
SFS-based demographic inference – using the site frequency spectrum
# |
Timepiece |
Tests |
Verified |
What it does |
|---|---|---|---|---|
X |
162 |
– |
Demographic inference from the SFS using moment equations. |
|
XI |
84 |
– |
Demographic inference from the SFS by solving the Wright-Fisher diffusion PDE. |
|
XII |
140 |
– |
Demographic inference from the SFS via coalescent tensor algebra. |
Genealogy and ARG inference – reconstructing ancestral histories
# |
Timepiece |
Tests |
Verified |
What it does |
|---|---|---|---|---|
III |
158 |
– |
The copying model: a haplotype as a mosaic of references. A gear inside many Timepieces. |
|
V |
120 |
– |
Bayesian ARG sampling with discretized time. SINGER’s predecessor. |
|
VI |
142 |
– |
Deterministic tree sequence inference at biobank scale. |
|
VII |
172 |
– |
Bayesian ARG sampling with continuous time and two-HMM architecture. |
|
VIII |
56 |
– |
Deterministic ARG inference at biobank scale with PBWT pre-filtering. |
|
XVII |
66 |
– |
Genome-wide genealogy estimation via asymmetric painting + MCMC branch lengths. |
Dating and selection – calibrating genealogies and detecting selection
# |
Timepiece |
Tests |
Verified |
What it does |
|---|---|---|---|---|
IX |
139 |
– |
Dates tree sequence nodes using the molecular clock. |
|
XV |
93 |
– |
Full-likelihood estimation of selection coefficients from gene trees and ancient DNA. |