Complex neuropsychiatric diseases such as schizophrenia require drugs that can target multiple G protein-coupled receptors (GPCRs) to modulate complex neuropsychiatric functions. Here, we report an automated system comprising a deep recurrent neural network (RNN) and a multitask deep neural network (MTDNN) to design and optimize multitargeted antipsychotic drugs. The system successfully generates novel molecule structures with desired multiple target activities, among which high-ranking compound 3 was synthesized, and demonstrated potent activities against dopamine D2, serotonin 5-HT1A and 5-HT2A receptors. Hit expansion based on the MTDNN was performed, 6 analogs of compound 3 were evaluated experimentally, among which compound 8 not only exhibited specific polypharmacology profiles but also showed antipsychotic effect in animal models with low potential for sedation and catalepsy, 2 highlighting their suitability for further preclinical studies. The approach can be an efficient tool for designing lead compounds with multitarget profiles to achieve the desired efficacy in the treatment of complex neuropsychiatric diseases.
KEY WORDSSchizophrenia; Multitargeted antipsychotic drugs; Recurrent neural network; Multitask deep neural network; Automated drug design Abbreviations: GPCRs, G protein-coupled receptors; RNN, deep recurrent neural network; MTDNN, multitask deep neural network; D2R, D2 receptors; 5-HT2AR, 5-HT2A receptors; 5-HT1AR, 5-HT1A receptors; EPS, Parkinson-like extrapyramidal symptoms; TD, tardive dyskinesia; HTS, high-throughput screening; AI, artificial intelligence; QSAR, quantitative structure-activity relationship; AEs, autoencoders; GANs, generative adversarial networks; RL, reinforcement learning; MW, molecular weight; logP, Wildman-Crippen partition coefficient; TPSA, total polar surface area; DNNs, deep neural networks; LSTM, long short-term memory; t-SNE, t-distributed stochastic neighbor embedding; ECFP4, extended connectivity fingerprint 4; SA, synthetic accessibility; QED, quantitative estimate of drug-likeness; R 2 , correlation coefficient; MAE, mean absolute error; RO5, Lipinski's rule of five; CNS, central nervous system; PCP, phencyclidine; NMDAR, N-methyl-D-aspartic acid receptor; SAR, structural-activity relationships; BPTT, backpropagation through time; ReLU, rectified linear unit; MSE, mean squared error