Detection of early-stage cancer using similarity-based classification of cell-free DNA fragmentation patterns

Blackjack classifies plasma cfDNA fragmentation patterns to detect early-stage cancer with minimum AUC 0.991 and sensitivity 89.8% at 98% specificity in stage I.

• Why it matters: Screening is limited by low adherence and missing tests for many cancers, so less invasive multi-cancer early detection using blood biomarkers is needed.
• What they did: Blackjack builds a position probability matrix of cfDNA fragmentation site sequence context, computes similarity features to healthy and cancer cohorts, then trains logistic regression; it is evaluated in cross-validation on two external whole-genome sequencing datasets.
• The result: Performance remains high across stage I (AUC 0.991, Sn 89.8%) and stage II (AUC 0.993, Sn 90.2%), across eight cancers (minimum AUC 0.985, minimum Sn 80.0%), and stays stable at ~0.24x simulated sequencing depth.

AI summary of the abstract · may contain errors · not medical advice · read the paper.