Artificial Intelligence

Harvard AI Model Identifies Genes and Drug Combinations to Reverse Disease in Cells

Published SEP 27 2025

Story

Image credit: Unsplash

The Facts

Harvard Medical School researchers have developed PDGrapher, an artificial intelligence model that identifies treatments capable of reversing disease states in cells by focusing on multiple drivers of disease rather than single protein targets.
Digital Watch Observatory
News Medical
The model utilizes graph neural networks to map relationships between genes, proteins and cellular pathways, and predicts optimal target combinations that restore healthy cell behavior from diseased states.
EurekAlert!
Medicalxpress
PDGrapher demonstrated superior performance when tested on 19 datasets spanning 11 cancer types, ranking correct therapeutic targets up to 35% higher than competing models and delivering results 25 times faster.
AI Insider
Nature
The tool successfully identified known drug targets, including KDR for non-small cell lung cancer and TOP2A enzyme for certain tumors, and also discovered additional candidates supported by emerging clinical evidence.
Inside Precision Medicine
Biotechniques
According to researchers, by uncovering the cause-and-effect biological drivers of disease, PDGrapher could help understand why certain drug combinations are effective — providing new biological insights that may accelerate further biomedical discoveries.
EurekAlert!
AI Insider
Research teams are currently applying PDGrapher to neurodegenerative conditions, including Parkinson's disease, Alzheimer's disease and X-linked Dystonia-Parkinsonism through collaborations with Massachusetts General Hospital.
News Medical
ZDNET

Sources Split

LEFT

CENTER

RIGHT

ESTAB
CRIT

CENTER

PRO
ESTAB

The Spin

Techno-skeptic narrative

AI models in drug discovery are fundamentally flawed black boxes that researchers shouldn't trust. These algorithms understand nothing about chemistry and make purely statistical correlations without a scientific basis. Most AI models excel only on retrospective benchmarks and rarely deliver actual prospective value in discovering real drug leads. As AI models often overfit datasets and struggle in real human systems, claims of restoring health in diseased cells may be premature without rigorous wet‑lab and clinical tests.

arXiv

PHYS

Sciencedirect

Techno-optimist narrative

Harvard's PDGrapher AI model is a breakthrough tool that actually reverses disease in cells by precisely targeting multiple pathways simultaneously. Unlike traditional single-target approaches, this advanced tool identifies the optimal gene combinations to restore healthy cell function. It has already validated known cancer targets and discovered new, previously overlooked ones with 35% greater accuracy than competing models. It offers a strong promise to significantly accelerate drug discovery — especially for complex diseases like cancer.