Scalable Log-ratio Lasso Regression Enhances Microbiome Feature Selection for Predictive Models

Identifying predictive biomarkers of patient outcomes from high-throughput microbiome data is of high interest in contemporary cancer research. We present FLORAL, an open-source computational tool to perform scalable log-ratio lasso regression modeling and microbial feature selection for continuous, binary, time-to-event, and competing risk outcomes. The proposed method adapts the augmented Lagrangian algorithm for a zero-sum constraint optimization problem while enabling a two-stage screening process for extended false-positive control. In extensive simulation studies, FLORAL achieved consistently better false-positive control compared to other lasso-based approaches and better variable selection [Formula: see text] score over popular differential abundance approaches. We demonstrate the practical utility of the proposed tool with a real data application on an allogeneic hematopoietic-cell transplantation cohort. The R package is available at https://github.com/vdblab/FLORAL.