Seonglae ChoPrima Mente
Prima Mente discovers novel biology to understand the human brain, to protect ourselves against disease and enhance ourselves in health.
https://www.primamente.com/Pleiades-July-2025/
Model interpretability can serve as a tool for discovering new disease biomarker hypotheses.
By interpreting an AI model (Pleiades) that detects Alzheimer's disease from cfDNA in blood, researchers found that DNA fragment length patterns contribute most significantly to predictions. Based on this interpretation, they created a simple, human-interpretable model using only fragment length features, which generalized well on independent data (AUROC≈0.78). Fragment length features may provide a stronger signal than the commonly used methylation-based biomarkers.
Using Interpretability to Identify a Novel Class of Alzheimer's Biomarkers
An AI model was trained to detect Alzheimer's from blood samples. We opened it up to understand how—and found that DNA fragment length patterns dominate its decision-making. We distilled this insight into a human-interpretable classifier that generalizes better than the biomarker classes previously reported in the literature when tested on an independent cohort.
https://www.goodfire.ai/research/interpretability-for-alzheimers-detection#
