Empirically calibrated simulations reveal the limits of phenotypic clustering algorithms for biodiversity assessment in data-scarce crops

PLOS One | https://doi.org/10.1371/journal.pone.0329254 December 17, 2025 2 / 14 Introduction Neglected and underutilized species (NUS) constitute a critical reservoir of agricul- tural biodiversity, essential for global food security, sustainable agriculture, and adap- tation to climate change and ecological instability [2,3]. Despite their resilience and adaptive potential, many NUS—such as fonio (Digitaria exilis), a cereal cultivated in ecologically marginal environments—remain underrepresented in genomic databases due to limited funding and molecular resources [1,4]. Consequently, scientists and breeders often rely on phenotypic data as a primary resource for exploring genetic structure, managing germplasm collections, and informing breeding strategies [5]. Yet, phenotypic analyses face fundamental challenges. These include subtle popu- lation differentiation and complex data distributions—characteristics often overlooked in methodological practice [6,7]. Fonio epitomizes this issue: historically vital to mar- ginalized West African communities, it thrives in agroecological niches unsuitable for major cereals [1,8]. In the absence of genomic profiling, phenotypic clustering is often the only available framework guiding biodiversity assessments and conservation interventions [9]. Crucially, the interpretability of phenotypic clustering depends on assumptions that are rarely met in NUS datasets—namely, that clusters are discrete, isotropic, and internally homogeneous [10,11]. In practice, phenotypic traits in NUS often exhibit

overlapping distributions, strong inter-trait correlations, heteroscedasticity, and low genetic differentiation (Pst ≈ 0.1–0.15), reflecting continuous gene flow and mixed mating systems [12–14]. Despite these biological complexities, most benchmarking studies assume ide- alized conditions. This disconnect raises concerns about the validity of widely used clustering metrics—such as Adjusted Rand Index (ARI) and Normalized Mutual Infor- mation (NMI)—which are sensitive to violations of assumptions like balanced sample sizes, distinct separation, and homogeneous variance [12,15]. While methodological critiques have emerged [13], few empirical studies have systematically quantified how these biases affect clustering outcomes in realistic biological scenarios. This gap is structural, not merely technical. Several popular algorithms illustrate this misalignment: • K-means and PAM assume spherical clusters with equal variances—conditions rarely met in empirical datasets [11]; • Ward’s method favors compact, globular clusters but collapses under trait-driven asymmetries [16]; • Gaussian Mixture Models (GMM) offer flexibility but still assume normality, often violated by real traits [17]; • DBSCAN accommodates arbitrarily shaped clusters but fails when densities vary across groups [18,19]. These algorithmic limitations have real-world consequences. Misclassification can result in the exclusion of locally adapted landraces or the loss of drought-resilient accessions—missed opportunities that are particularly costly in resource-limited distribution, and reproduction in any medium,

provided the original author and source are credited. Data availability statement: All empirical parameters used for simulation calibration were extracted from [1]. Agromorphological charac- terization revealed three phenotypic groups in a region-wide germplasm of fonio (Digitaria exilis (Kippist) Stapf) from West Africa. Agronomy, 10(11), 1653., whose full dataset is publicly available within the article and its supplemen- tary tables. All relevant data are within the manuscript and its Supporting information files. In addition, the complete R script used to simulate phenotypic datasets, apply clustering algorithms, and compute evaluation metrics is publicly available on Zenodo: https://doi. org/10.5281/zenodo.15877863. This ensures full reproducibility of the results presented. Funding: The author(s) received no specific funding for this work. Competing interests: The author has declared that no competing interests exist.