LoginSign Up
← Back to all sparks
P

pedbuildr

ANALYTICS
Velocity0.0

Pedigree Reconstruction

github.com/magnusdv/pedbuildr ↗

pedbuildr reconstructs pedigrees from DNA, and it just got much faster at the search.

pedigree reconstructionforensic geneticsparallel computingperformancer packages
◆Current state
pedbuildr infers pedigree structure from marker data by scoring candidate pedigrees against likelihoods. Version 0.4.0 moved reconstruct() to mirai for parallel likelihood computation, deprecated the old numCores argument, and picked up the improved loop breaking from pedtools and pedprobr, which lets many complex looped pedigrees succeed where they previously failed. buildPeds() also got significantly faster in its default configuration where mating between lineally related individuals is disallowed.
◆Where it's heading
The package is bounded by two costs: how many candidate pedigrees it enumerates and how expensive each likelihood is. The recent release attacks both, parallelizing the likelihoods and speeding up enumeration in the common case. The earlier 0.3.0 release worked on the other end, adding inbreeding limits and a proper result class so the output of a large search is manageable. Releases are infrequent, roughly three years apart in this window.
◆Prediction
Expect the candidate generation side to receive the same attention the likelihood side just did, since search space size is the remaining bound on what pedbuildr can reconstruct.

◆Recent moves

  1. 1mo ago

    mirai parallelism and faster pedigree enumeration

    reconstruct() computes likelihoods in parallel through mirai daemons, replacing the deprecated numCores argument, and improved loop breaking in pedtools and pedprobr means many complex looped pedigrees now succeed instead of failing. buildPeds() is also significantly faster in the default no-lineal-inbreeding case.

    View source ↗
  2. 2y ago

    pedCollection class and an inbreeding ceiling

    buildPeds() returns a pedCollection object with print, plot, and subsetting methods, and gains maxInbreeding to cap the tolerated inbreeding coefficient, defaulting to the first-cousin level appropriate for human pedigrees. The Tutankhamun dataset is added as a worked example, and the main functions get a general speedup.

    View source ↗