henhaobio
Sign in

bioRxivBioinformaticsPreprint

BITE: an R package for biodiversity analyses

Marco MilanesiiD, Stefano CapomaccioiD, Elia VajanaiD, Lorenzo BombaiD, José Fernando Garcia, Paolo Ajmone‐MarsaniD, Licia ColliiD

PDFDOICode
In-depth full-text digest available · Jump to digest ↓

Abstract

Nowadays, molecular data analyses for biodiversity studies often require advanced bioinformatics skills, preventing many life scientists from analyzing their own data autonomously. BITE R package provides complete and user-friendly functions to handle SNP data and third-party software results (i.e. Admixture, TreeMix), facilitating their visualization, interpretation and use. Furthermore, BITE implements additional useful procedures, such as representative sampling and bootstrap for TreeMix, filling the gap in existing biodiversity data analysis tools.\n\nAvailabilityhttps://github.com/marcomilanesi/BITE

In-depth summary

Claude Opus 5.5 · max effort · 8 Nov 2026 · from the full text of version 1 (29 Aug 2017) · CC BY-NC-ND 4.0

BITE, an R package, wraps SNP quality control, Kolmogorov-Smirnov-checked representative subsampling, color-consistent ancestry plots, and bootstrap-supported TreeMix gene-flow trees for biodiversity genomics.

As SNP datasets reach big-data scale, biologists moving into bioinformatics must spend time learning to program, choose among a crowded field of tools, and stitch separate analyses into one pipeline. BITE gathers such analyses in R, adding one-call GenABEL-based quality control, a routine seeking subsamples that mirror the full dataset's genetic structure, ancestry plots with stable colors across cluster numbers (K), and bootstrap node support for TreeMix.

Key results

  1. ClaimedBITE bundles SNP-based biodiversity workflows into R functions: transposed PLINK input processed via GenABEL, single-call quality control, marker- and individual-level summary statistics with plots, and outputs formatted for third-party software.

    bite.qc merges GenABEL's check.marker steps into one call and writes retained and excluded SNPs and individuals as PLINK files; bite.biostat reports marker and individual statistics with accompanying graphics. No runtime, usability testing, or comparison with other packages is reported. · 1 Introduction; 2.1 Genomic data handling

  2. SuggestedA TreeMix bootstrap mode, newly added by BITE, attaches support values to tree nodes: replicate trees at the selected migration count are merged into a PHYLIP consensus tree, on which TreeMix then re-fits that number of migration edges.

    treemix.scripts emits BASH scripts (parallelized with GNU parallel) for conventional or bootstrap runs; newick.split maps bootstrap values onto Newick splits and treemix.bootstrap plots them. Figure 1B shows node support from 100 bootstrap replicates on the human HapMap trial data; accuracy of support values is not evaluated. · 2.3 Gene flow analyses; Figure legends

  3. ClaimedThe representative.sample function seeks a subset of individuals, of user-chosen size, whose genetic structure matches that of the original dataset, targeting bias from unequal population sample sizes where the representativeness of random subsets is contested.

    Multidimensional scaling of pairwise identity-by-state distance matrices is run for original and subsampled sets; on each of p user-set dimensions a two-sample Kolmogorov-Smirnov test compares score distributions, and the subset is accepted when no dimension differs significantly. No simulation or validation is shown. · 2.1 Genomic data handling

  4. SuggestedMembership coefficients from Admixture, FastSTRUCTURE and sNMF are drawn with each genomic component keeping one color across all K, as standard or full/half circular Circos-style bar plots, plus a helper plotting model-selection metrics to choose K.

    Implemented as membercoef.plot, membercoef.circos (modified RCircos code) and membercoef.cv. Figure 1A displays Admixture coefficients for the human trial data at K = 2, 3, 4, 7 in circular form; the claimed gain in interpretability is not quantified. · 2.2 Graphical representation of membership…; Figure legends

  5. ClaimedFurther TreeMix helpers report model fit as the fraction of explained variance in between-population relatedness, draw heatmaps of genetic drift among populations, and bundle Pickrell and Pritchard's original plotting functions.

    Functions treemix.fit and treemix.drift, plus the Pickrell and Pritchard (2012) R code. bite.trialdata ships TreeMix outputs for 1,011 HapMap Phase 3 individuals from 11 populations in conventional (m=0-5) and bootstrap (m=4) mode, but no fit values or drift estimates are reported in the text. · 2.3 Gene flow analyses; 3 Example

Methods, data and reproducibility

R package built on GenABEL that takes transposed PLINK SNP data through quality control, subsampling checked by multidimensional scaling of identity-by-state distances plus per-dimension Kolmogorov-Smirnov tests, summary statistics, ancestry plotting for Admixture/FastSTRUCTURE/sNMF outputs, and bootstrapped TreeMix runs with PHYLIP consensus trees.

  • HapMap Project Phase 3 subset: 11 human populations with 20,000 randomly chosen SNPs on chromosome 2, provided as binary PLINK files via bite.trialdata · human · n = 1,011 subjects; 11 populations; 20,000 SNPs
  • Precomputed Admixture, FastSTRUCTURE and sNMF membership-coefficient outputs for the HapMap trial subset (K=2 to K=10) · human · n = 1,011 subjects
  • Precomputed TreeMix outputs for the HapMap trial populations, from conventional runs (m=0-5) and bootstrap mode (m=4) · human · n = 11 populations
Methods & tools
BITE R package, GenABEL (check.marker), PLINK formats (transposed and binary), Admixture, FastSTRUCTURE, sNMF, TreeMix, PHYLIP (consensus trees), GNU parallel, RCircos / Circos-style plotting, Multidimensional scaling of identity-by-state distances, Two-sample Kolmogorov-Smirnov test
Software & environment
R package built on GenABEL functions; TreeMix runs are scripted in BASH, and bootstrap consensus trees are built with PHYLIP; no software versions are given.
Assay conditions & compute
TreeMix BASH scripts use GNU parallel to exploit multi-threaded machines; no specific hardware, runtime, or memory figures are reported.
Reproducibility gaps
No package, R, or dependency versions or archived release are stated; trial data are available only through bite.trialdata inside the package, without a HapMap source accession.
Limitations (3)
  • Unequal population sample sizes can bias biodiversity analyses, and whether a subsample represents its source population remains debated with no universally accepted solution; representative.sample is presented as an attempt rather than a settled criterion. · stated by the authors
  • No quantitative evaluation is given: representative.sample is neither benchmarked against random subsampling nor tested in simulations, bootstrap support accuracy is untested, no runtimes address the big-data motivation, and the only demonstration is one human HapMap subset. · noted in this summary
  • The Kolmogorov-Smirnov acceptance rule leaves the significance level, multiple-testing correction across the p dimensions, and the way candidate subsets are generated unspecified; non-significant tests also do not prove distributional equivalence, particularly for small subsets. · noted in this summary

Read this if you analyze SNP genotypes for population structure or gene flow in R and want one toolkit covering quality control, color-consistent ancestry plots across K, and bootstrap node support for TreeMix trees.

Citations
211

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISNitta '22Bilderbeek '18Jong '21Zizka '15Simmons '18Su '19
61% semantic matchEvolutionary Biology
J. Nitta, S. W. Laffan, B. D. Mishler+1 more·bioRxiv·Oct 6, 2022