ALFAssay: A feed-forward neural network for quantitative fragmentomics-based ctDNA profiling in breast cancer.
A machine-learning blood test detects circulating cancer DNA in breast cancer patients with high accuracy, even when traditional genetic markers are sparse.
ALFAssay uses 408-dimensional fragmentation feature vectors (short-to-total fragment ratios in 5 Mb bins, accounting for coverage effects) from shallow WGS to quantify ctDNA fraction in breast cancer (896 plasma samples: HR+/HER2-, TNBC, early and metastatic stages, plus healthy controls). The model achieves sensitivity 0.87, specificity 0.94 for ctDNA detection and provides independent prognostic value for PFS stratification—particularly valuable for tumors with low copy-number aberration burden where existing shallow WGS tools perform poorly.
What the study was
- Study design
- model_development_and_validation
- Population
- Breast cancer patients with plasma samples for shallow WGS (n=896; HR+/HER2-, TNBC, early and metastatic stages, plus healthy controls)
- Sample size
- 896
- Category
- Early Detection
- Maturity
- Exploratory
- Journal
- PLoS Comput Biol
Why it surfaced
Validated neural network tool (PLoS Comput Biol, n=896 breast cancer plasma samples) for ctDNA quantification from fragmentomics via shallow WGS, filling a practical gap for low-CNA-burden tumors and complementing existing multi-modal liquid biopsy workflows with prognostic stratification utility.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.