shaai schreef op 27 december 2025 15:46:
En bijv 4 weken terug:
Foundational model achieves 45% hit rate vs. <5% industry standard in Parkinson's target discovery
Excited to share results from our collaboration with UCL Queen Square Institute of Neurology that demonstrate the power of combining AI predictions with rigorous experimental validation.
Our foundational model prioritized 198 candidate genes associated with lysosomal and mitochondrial dysfunction in Parkinson's disease. Iria Trasobares Magdalena and the UCL team then experimentally validated these predictions using dual high-throughput siRNA screens measuring PINK1/Parkin-dependent mitophagy.
Results:
? ~19% hit rate in early pathway activation (pUb(Ser65) readout)
? ~45% hit rate in late-stage mitophagy (mitochondrial-lysosomal delivery)
? Identification of a novel negative regulator validated across multiple readouts
This 9-fold enrichment over typical screening hit rates shows how computational biology and experimental neuroscience can work together to accelerate target discovery.
Read Iria Trasobares Magdalena's excellent blog post detailing the experimental work and findings:
lnkd.in/egKvqrmj