Prediction of Compound Bioactivities Using Heat-Diffusion Equation

• Published in: Patterns (New York, N.Y.) N.Y.), 2020-12, Vol.1 (9), p.100140-100140, Article 100140
• Main Authors: Hidaka, Tadashi, Imamura, Keiko, Hioki, Takeshi, Takagi, Terufumi, Giga, Yoshikazu, Giga, Mi-Ho, Nishimura, Yoshiteru, Kawahara, Yoshinobu, Hayashi, Satoru, Niki, Takeshi, Fushimi, Makoto, Inoue, Haruhisa
• Format: Article
• Language: English
• Subjects: ALS; chemotype; compound screening; drug discovery; heat-diffusion equation; iPSC panel; machine learning; phenotypic screening; prediction
• ISSN: 2666-3899; 2666-3899
• DOI: 10.1016/j.patter.2020.100140

Machine learning is expected to improve low throughput and high assay cost in cell-based phenotypic screening. However, it is still a challenge to apply machine learning to achieving sufficiently complex phenotypic screening due to imbalanced datasets, non-linear prediction, and unpredictability of new chemotypes. Here, we developed a prediction model based on the heat-diffusion equation (PM-HDE) to address this issue. The algorithm was verified as feasible for virtual compound screening using biotest data of 946 assay systems registered with PubChem. PM-HDE was then applied to actual screening. Based on supervised learning of the data of about 50,000 compounds from biological phenotypic screening with motor neurons derived from ALS-patient-induced pluripotent stem cells, virtual screening of >1.6 million compounds was implemented. We confirmed that PM-HDE enriched the hit compounds and identified new chemotypes. This prediction model could overcome the inflexibility in machine learning, and our approach could provide a novel platform for drug discovery. [Display omitted] •Prediction model based on heat-diffusion equation (PM-HDE) was constructed•PM-HDE succeeded in increasing the hit ratio and identifying potent compounds•PM-HDE discovered new chemotypes in compound evaluation with an ALS-patient iPSC panel•PM-HDE could represent an algorithm for future drug discovery with AI There remain many intractable diseases with no treatment available, including amyotrophic lateral sclerosis (ALS), for which the development of a cure is crucial. However, compound screening for drug development demands time, energy, and cost, and therefore artificial intelligence (AI) is expected to improve the efficiency of drug discovery. We built a novel machine-learning algorithm to predict hit compounds in compound screening using the heat-diffusion equation (HDE). This prediction model harbors the potential to solve issues that have been challenging for conventional machine learning and to exhibit accurate performance leading to the discovery of new drugs. In fact, the HDE model predicted hits with new chemotypes among millions of compounds for ALS therapeutics using a panel of large numbers of ALS patient-derived induced pluripotent stem cell models (ALS-patient iPSC panel). This algorithm could contribute to the acceleration and development of future drug discoveries using AI. Compound screening is a useful tool for discovering new candidate drugs. However, it is still a maj