GCTAg_benchmark

Scripts for benchmarking GCTAg against the original GCTA.

All GCTAg flags, options and release notes live in the GCTAg repository. This repo only contains the tooling used to run, compare and plot benchmarks.

Layout

Directory	Purpose
benchmarks/	Scripts that run GCTA and GCTAg and record elapsed time and peak RSS
plotting/	Scripts that turn benchmark outputs into figures
scripts/	Supporting tools: simulation, format conversion, etc.

benchmarks/

Drivers for timing and memory comparisons between GCTA and GCTAg across the GRM, PCA, REML and MLMA stages. The main entry point is benchmarks/run_benchmark.sh, so can run the e.g. end-to-end benchmark as benchmarks/run_benchmark.sh End_to_end, which will set up and execute End_to_end.sh

plotting/

Scripts that read the benchmark outputs and produce the comparison figures (runtime, memory, accuracy vs. GCTA).

• plot_benchmarks.py: plots runtime and peak RSS from the output of benchmarking scripts
• plot_mlma.py: plots either Manhattan plots or MLMA concordance between multiple ".mlma" files
• plot_reml_trajectory.py: plots the per-round variance components estimation

scripts/

• simulate_population.R: simulates cattle population with AlphaSim
• BED_to_MACH.cpp: converts PLINK bfiles to MACH format (see below)
• generate_data_splits.sh: creates subsets of genotype data from PLINK bfiles

MaCH conversion

Compile the converter with

g++-16 -O3 -fopenmp script/BED_to_MACH.cpp -o bed_to_mach -lz

then convert PLINK bfiles to MACH format with

bed_to_mach --bfile kbhm --out MACH --threads 2