Principal Component Analyses (PCA)-based findings in population genetic studies are highly biased and must be reevaluated

Abstract

Principal Component Analysis (PCA) is a multivariate analysis that reduces the complexity of datasets while preserving data covariance. The outcome can be visualized on colorful scatterplots, ideally with only a minimal loss of information. PCA applications, implemented in well-cited packages like EIGENSOFT and PLINK, are extensively used as the foremost analyses in population genetics and related fields (e.g., animal and plant or medical genetics). PCA outcomes are used to shape study design, identify, and characterize individuals and populations, and draw historical and ethnobiological conclusions on origins, evolution, dispersion, and relatedness. The replicability crisis in science has prompted us to evaluate whether PCA results are reliable, robust, and replicable. We analyzed twelve common test cases using an intuitive color-based model alongside human population data. We demonstrate that PCA results can be artifacts of the data and can be easily manipulated to generate desired outcomes. PCA adjustment also yielded unfavorable outcomes in association studies. PCA results may not be reliable, robust, or replicable as the field assumes. Our findings raise concerns about the validity of results reported in the population genetics literature and related fields that place a disproportionate reliance upon PCA outcomes and the insights derived from them. We conclude that PCA may have a biasing role in genetic investigations and that 32,000-216,000 genetic studies should be reevaluated. An alternative mixed-admixture population genetic model is discussed.

Constructing Certainty in Machine Learning: On the performativity of testing and its hold on the future

“…Current incentive structures encourage simplified narratives of progress based on neat metrics that can be tuned to make one's algorithms look 5 Similar issues on downplaying judgment have been found in studies using classical statistical analysis based on pvalues and confidence intervals (Ioannidis 2005;Berman et al 2018). A recent study also illustrates how assumed mechanistic objectivity in Principle Component Analysis, a method used both in ML and statistical analysis, may require thousands of population genetics studies to be reevaluated (Elhaik 2022). The focus of this paper is on ML, but these shared issues illustrate how some parts of the analysis are also relevant to other fields.…”

Decision Tree Ensembles Utilizing Multivariate Splits Are Effective at Investigating Beta-Diversity in Medically Relevant 16S Amplicon Sequencing Data

“…When moving beyond toy data, there exists a considerable amount of work advocating for the adoption of newer graph-based methods for the analysis of biological data. For example, Elhaik (2022) demonstrates that the validity of the results produced from one commonly used linear dimensionality reduction algorithm, PCA, can be easily called into question due to artifacts within the data and the ease by which it can be manipulated to produce favorable results (6). In addition, there is an ongoing debate around how useful a projection is when the amount of explained variance for the first few axes is very low (arbitrarily defined as less than 60%).…”

“…A considerable amount of work has demonstrated that many genomic, transcriptomic, and metagenomic datasets are better understood if the underlying geometry of the input space is considered. This better understanding can be reached since these datasets do not often meet the "Assumption of Linearity" used by methods such as Non-Negative Matrix Factorization, Principal Components Analysis and its generalization, Principal Coordinates Analysis (5)(6)(7)(8). Finally, these methods rarely explore correlations and other dependencies between taxa.…”

Asymmetrical lineage introgression and recombination in populations of Aspergillus flavus: Implications for biological control

“…Population structure was also examined using principal component analysis (PCA) implemented in EIGENSOFT SmartPCA v18140 [ 81 ]. A recent study has shown that PCA can yield highly biased results and should not be used as a first hypothesis generator in population genetic analyses [ 82 ]. Here, PCA was used to corroborate k cluster estimation that was based primarily on phylogenetic reconstruction methods assuming bifurcation or reticulation, and admixture models to characterize ancestral genetic structure.…”