English, N. J. | Search bioRxiv

bioRxiv2025

Explainable machine learning for health disparities: type 2 diabetes in the All of Us research program

Type 2 diabetes (T2D) is a disease with high morbidity and mortality and a disproportionate impact on minority groups. Machine learning (ML) is increasingly used to characterize T2D risk factors; however, it has not been used to study T2D health disparities. Our objective was to use explainable ML methods to discover and characterize T2D health disparity risk factors. We applied SHapley Additive exPlanations (SHAP), a new class of explainable ML methods that provide interpretability to ML classifiers, to this end. ML classifiers were used to model T2D risk within and between self-identified race and ethnicity (SIRE) groups, and SHAP values were calculated to quantify the effect of T2D risk factors. We then stratified SHAP values by SIRE to quantify the effect of T2D risk factors on prevalence differences between groups. We found that ML classifiers (random forest, lightGBM, and XGBoost) accurately modeled T2D risk and recaptured the observed prevalence differences between SIRE groups. SHAP analysis showed the top seven most important T2D risk factors for all SIRE groups were the same, with the order of importance for features differing between groups. SHAP values stratified by SIRE showed that income, waist circumference, and education best explain the higher prevalence of T2D in the Black or African American group, compared to the White group, whereas income, education and triglycerides best explain the higher prevalence of T2D in the Hispanic or Latino group. This study demonstrates that explainable ML can be used to elucidate health disparity risk factors and quantify their group-specific effects. Author SummaryWhile machine learning (ML) methods hold great promise for epidemiological studies, their practical utility is limited by interpretability. Increasingly complex ML models are great at predicting disease risk, but how they arrive at a given prediction is often obscured by model complexity. Explainable ML is an emerging discipline that seeks to render ML models more transparent by elucidating how and why input features contribute to output predictions. This study reports a novel application of explainable ML to epidemiology, focusing on type 2 diabetes (T2D) as a paradigm of health disparities. We found that ML classifiers were able to accurately model T2D disparities, for a large cohort of Black, Hispanic, and White Americans, and explainable ML revealed which risk factors contributed to the observed disparities and how. The results demonstrate that explainable ML can be a powerful tool for the discovery and characterization of health disparity risk factors.

bioinformatics↗

bioRxiv2023

Model-based inference of a plant-specific dual role for HOPS in regulating guard cell vacuole fusion

Stomata are the pores on a leaf surface that regulate gas exchange. Each stoma consists of two guard cells whose movements regulate pore opening and thereby control CO2 fixation and water loss. Guard cell movements depend in part on the remodeling of vacuoles, which have been observed to change from a highly fragmented state to a fused morphology during stomata opening. This change in morphology requires a membrane fusion mechanism that responds rapidly to environmental signals, allowing plants to respond to diurnal and stress cues. With guard cell vacuoles being both large and responsive to external signals, stomata represent a unique system in which to delineate mechanisms of membrane fusion. Fusion of vacuole membranes is a highly conserved process in eukaryotes, with key roles played by two multi-subunit complexes: HOPS (homotypic fusion and vacuolar protein sorting) and SNARE (soluble NSF attachment protein receptor). HOPS is a vacuole tethering factor that is thought to chaperone SNAREs from apposing vacuole membranes into a fusion-competent complex capable of rearranging membranes. To resolve a counter-intuitive observation regarding the role of HOPS in regulating plant vacuole morphology, we derived a quantitative model of vacuole fusion dynamics and used it to generate testable predictions about HOPS-SNARE interactions. We derived our model by applying simulation-based inference to integrate prior knowledge about molecular interactions with limited, qualitative observations of emergent vacuole phenotypes. By constraining the model parameters to yield the emergent outcomes observed for stoma opening - as induced by two distinct chemical treatments - we predicted a dual role for HOPS and identified a stalled form of the SNARE complex that differs from phenomena reported in yeast. We predict that HOPS has contradictory actions at different points in the fusion signaling pathway, promoting the formation of SNARE complexes, but limiting their activity. Author summaryPlants "breathe" through pores in their leaves where each pore is formed by two specialized cells called guard cells. To open these pores, guard cells change in volume. This volume change is controlled by water-filled organelles called vacuoles that morph from multiple small entities to a few large ones capable of taking up more water to reshape the cell. Specialized proteins in vacuole membranes make this change happen by pulling vacuoles together until they fuse. Some of these proteins reside in membranes, but others must be drawn to the membrane from the cells cytoplasm. Specific lipid molecules in the membrane play an important role in recruiting those proteins to the vacuole membrane. We previously made an unexpected finding that removing this lipid induces plant vacuole fusion. To make sense of this observation, we used a mathematical model to piece together our knowledge of the proteins involved in this process and what we know about the chemical treatments that cause vacuoles to morph. Using computer simulations, we uncovered new rules about how molecules interact in membranes to accomplish the task of vacuole fusion in plants. We think the rules uncovered through mathematical modeling allow plants to respond quickly to environmental cues.

systems biology↗