Interpreting the flock algorithm from a statistical perspective.

Area of Science:

• Population genetics
• Computational biology
• Bioinformatics

Background:

• The flock algorithm (Duchesne & Turgeon 2009) is interpreted as an estimation procedure.
• It is based on a model similar to the structure model (Pritchard et al. 2000) but without admixture or correlated allele frequency priors.
• Flock uses simulated annealing for maximum a posteriori estimation, unlike structure's MCMC.

Discussion:

• The flockture program, a modified version of structure using the flock algorithm, processes data ~200 times faster.
• Both flockture and structure were evaluated on simulated datasets with varying population differentiation for microsatellite and SNP genotypes.
• Performance was assessed based on the ability to cluster individuals into their correct subpopulations.

Key Insights:

• Flockture yields results comparable to structure, though with increased run-to-run variability.
• Flockture demonstrated superior performance over structure for SNP genotypes with moderate population differentiation (FST=0.022-0.032).
• Structure outperformed flockture under conditions of low population differentiation for both marker types.

Outlook:

• Understanding flock's algorithm as a specific case of the structure model can clarify its output and behavior.
• The flock algorithm's reliance on 'plateau record' inference rules may not be beneficial for large datasets.
• Further research could explore optimizing flockture's performance or its applicability to diverse population genetic scenarios.