Predicting Cancer Gene Expression from Histology Images with SEQUOIA

A pioneering study introduces SEQUOIA, a linearized transformer model designed to predict cancer transcriptomic profiles directly from whole slide images (WSIs). By analyzing 7,584 tumor samples across 16 cancer types and validating results on two independent cohorts of 1,368 tumors, SEQUOIA accurately identified genes linked to key cancer processes, such as inflammation, cell cycles, and metabolism. Beyond genetic prediction, SEQUOIA demonstrated potential in breast cancer recurrence risk stratification and spatial resolution of gene expression at loco-regional levels. This innovative approach leverages deep learning to extract clinically valuable information from WSIs, paving the way for cost-effective, personalized cancer management.