{"ok":true,"nota":{"id":"37f46bf9859e66e8","slug":"un-programa-de-inteligencia-artificial-identifica-mejor-que-genes-sirv-37f46b","titulo":"Un programa de inteligencia artificial identifica mejor qué genes sirven como blanco para medicamentos","resumen":"Desarrollaron un modelo de inteligencia artificial que analiza variantes genéticas humanas para priorizar genes candidatos a tratamientos. En pruebas difíciles, el sistema superó a métodos previos en aciertos entre sus primeras 20 predicciones.","fuente":"biorxiv","fuenteNombre":"bioRxiv · preprints","color":"#be2628","autor":"Wenteler, A., Neduva, V., Genes & Health Research Team,, Cabrera, C. P., Wei, W., Barnes, M. R.","url":"https://www.biorxiv.org/content/10.64898/2026.10.16.759653v1?rss=1","imagen":"/hoy/portada/37f46bf9859e66e8","tema":"tecnologia","subtemas":["inteligencia artificial","genetica","desarrollo de medicamentos","variantes genéticas","aprendizaje automatico"],"entidades":["Varformer","DrugnomeAI"],"formato":"dato","entretenimiento":false,"pubTs":"2026-10-22T00:00:00.000Z","indexTs":"2026-10-22T21:54:18.823Z","hash":"feb573a272539e9c5e794efd624246e962fa0e972f720f8be1f11379458280bc","tsa":{"time":"Oct 22 21:55:19 2026 GMT","provider":"freetsa.org","sobre":"sello"},"stats":{"views":0,"likes":0,"clicks":0,"dwell":0},"tipo":"normal","registro":"hallazgo","porQueImporta":"Encontrar un blanco para un medicamento puede demorar años y costar fortunas, así que una herramienta que acierte mejor reduce la cantidad de intentos perdidos. Para la salud pública, esto podría acelerar el desarrollo de tratamientos para enfermedades genéticas que hoy tienen pocas opciones. Sin embargo, el modelo se entrenó con objetivos ya validados en cuatro grupos de ancestros, y su rendimiento cae cuando la tarea se vuelve más difícil.","quienFirma":"bioRxiv · preprints","confianza":"preprint","consulta":null},"sello":{"hash":"feb573a272539e9c5e794efd624246e962fa0e972f720f8be1f11379458280bc","sello":"270ae771d09657267a4363277fde3fc372296b33916c340a3f34f76125306e55","beacon":{"drand":{"cadena":"8990e7a9aaed2ffed73dbd7092123d6f289930540d7651336225dc172e51b2ce","ronda":6575836,"aleatorio":"a8739c6f28d7e03ed191a881119536d56c373334288c8dc3d26ef33eb54db44f","tiempo":"2026-10-22T21:55:00.000Z"},"nist":{"cadena":2,"pulso":1995117,"salida":"932C769A8B73E2CC53835512DA1C7CF78DC4B06757B136D8DCC28C89B61C25ED52E0DD74B99E06F83F83805D89C905466FC5BF54DA4D037DEA5F5312DBA4C659","tiempo":"2026-10-22T21:54:00.000Z"},"noAntes":"2026-10-22T21:55:00.000Z"},"verificacion":{"drand":true,"nist":true},"tsa":{"time":"Oct 22 21:55:19 2026 GMT","provider":"freetsa.org","sobre":"sello"},"original":{"titulo":"Population-scale drug target and variant prioritization with Varformer","descripcion":"Human genetic variation is a powerful driver of therapeutic target prioritization, yet current models struggle to aggregate variable numbers of variants per gene and are bottlenecked by the scarcity of validated targets. We present Varformer, a multimodal transformer that uses cross-modal attention to integrate contextual representations of missense variants with gene-level features. Trained via positive-unlabeled learning on clinically validated targets across four genetic ancestry groups, Varformer achieves a precision@20 of 0.95, 0.86, and 0.48 across three benchmarks of increasing difficulty. It surpasses a logistic regression ablation by 20% and DrugnomeAI by 23% on the hardest benchmar","pubDate":"2026-10-22","url":"https://www.biorxiv.org/content/10.64898/2026.10.16.759653v1?rss=1"},"indexado":"2026-10-22T21:54:18.823Z"}}