Psilocybin therapy for treatment resistant depression: prediction of clinical outcome by natural language processing

• DOI: 10.1007/s00213-023-06432-5
• Publication Date: 2023-08-22T06:01:50Z
• Authors: Robert F. Dougherty; Patrick Clarke; Merve Atli; Joanna Kuc; Danielle Schlosser; Boadie W. Dunlop; David J. Hellerstein; Scott T. Aaronson; Sidney Zisook; Allan H. Young; Robin Carhart-Harris; Guy M. Goodwin; Gregory A. Ryslik
• Keywords: Psilocybin

Therapeutic administration of psychedelics has shown significant potential in historical accounts and recent clinical trials the treatment depression other mood disorders. A randomized double-blind phase-IIb study demonstrated safety efficacy COMP360, COMPASS Pathways' proprietary synthetic formulation psilocybin, participants with treatment-resistant depression.While results are promising, works for a portion population early prediction outcome is key objective as it would allow identification those likely to require alternative treatment.Transcripts were made from audio recordings psychological support session between participant therapist 1 day post COMP360 administration. zero-shot machine learning classifier based on BART large language model was used compute two-dimensional sentiment (valence arousal) transcript. These scores, combined Emotional Breakthrough Index (EBI) arm predict measured by MADRS scores. (Code data available at https://github.com/compasspathways/Sentiment2D .) RESULTS: Two multinomial logistic regression models fit responder status week 3 through 12. Cross-validation these resulted 85% 88% accuracy AUC values 85%.A algorithm using NLP EBI accurately predicts long-term patient response, allowing rapid prognostication personalized response psilocybin insight into therapeutic optimization. Further research required understand if earlier stages process hold similar predictive power.