The AI Tool Hunting for Orphan Drug Targets

Insilico Medicine's TargetPro-TargetBench framework boosts rare disease target accuracy, reshaping how orphan drugs are found

10 Jun 2026

A long-standing bottleneck in orphan drug development may be cracking open. On April 16, 2026, Insilico Medicine announced a validated AI framework that attacks one of the most persistent causes of clinical failure in rare disease research: weak biological targets. Nearly nine in ten drug candidates in the field never reach patients.

Central to the advance is the TargetPro-TargetBench system. TargetPro is a disease-specific machine learning model trained across 38 disease categories spanning oncology, neurological, fibrotic, immune-related, and metabolic conditions, and integrates 22 distinct omics and text-based data scores from Insilico's PandaOmics platform. In performance tests, it retrieved 71.6 percent of known clinical targets, representing a 1.7- to 5.5-fold improvement over leading large language model approaches, according to company statements.

Expanded versions of both tools now widen that scope considerably. TargetPro 2.0 and TargetBench 2.0 extend disease coverage from 38 to 100 indications across 10 therapeutic areas, adding cardiovascular, ophthalmological, reproductive, and mental health disorders. For rare diseases, where patient populations are too small to sustain conventional trial-and-error discovery, that breadth matters. Predicted novel targets showed 95.7 percent structure availability and 86.5 percent druggability, alongside meaningful repurposing potential for existing compounds, analysts noted.

Benchmarking has historically lagged behind prediction in AI drug discovery, creating a credibility gap that slows institutional adoption. TargetBench addresses this by giving researchers a standardized tool to evaluate and compare target identification models against a common evidence base. The framework, validated in Scientific Reports, offers a shared evaluation standard that other organizations can apply independently.

For the rare disease community, the practical implications are tangible. Faster, more reliable target identification compresses the front end of drug development, where costs and timelines have traditionally been most punishing. As platforms of this kind mature, the results could shape treatment pathways and regulatory expectations in the years ahead.