The prediction of compound-protein interaction (CPI) is a critical technological tool for discovering lead compounds and drug repurposing during the process of drug development.In recent years, deep learning has been widely used in CPI research, which has accelerated the development of CPI prediction in drug discovery.This review focuses on feature-based CPI prediction models.First, we described the datasets, as well as typical feature representation methods commonly used for compounds and proteins in CPI prediction.Based on the critical problems in modeling, we discussed models for CPI prediction from two perspectives: multimodal features and attention mechanisms.Then, the performance of 12 selected models was evaluated on 3 benchmark datasets for both classification and regression tasks.Finally, the review summarizes the existing challenges in this field and prospects for future directions.We believe that this investigation will provide some reference and insight for further research on CPI prediction.

Ulcerative colitis(UC)is a common disease of the digestive system.Phosphatidylinositol-3-kinase(PI3K)/synuclein/threonine kinase(AKT)is closely related to cell survival,apoptosis,inflammation and other biological processes,and the expression levels of PI3K and AKT significantly increase during the course of UC,with accelerated apoptosis,improved inflammation,and damaged intestinal mucosal barrier function.In recent years,a large number of basic and clinical trials have been conducted on PI3K/AKT signaling pathway in TCM,and the results indicate that PI3K/AKT signaling pathway is expected to be an important potential target for UC treatment.This article analyzed the mechanism of the regulation of PI3K/AKT signaling pathway in TCM from monomer,extract,compound and acupuncture,and suggested that the regulation of this signaling pathway is of great significance for the prevention and treatment of UC,and provide reference for drug development.

Kirsten rat sarcoma viral oncogene homolog(KRAS)gene is one of the most commonly mutated oncogenes.It has been found that KRAS inhibitors have the potential therapeutic effect on cancer patients with this gene mutation.In this study,machine learning was applied to develop a QSAR(quantitative structure-activity relationship)model for KRAS small molecule inhibitors.A total of 1857data points of IC50 and SMILES(simplified molecular input line entry system)for KRAS inhibitors were collected from three databases:ChEMBL,BindingDB,and PubChem.And nine different classifiers were constructed using three different feature screening methods combined with three machine learning models,namely,random forest,support vector machine,and extreme gradient boosting machine.The results showed that the SVM model combined with mutual information feature selection exhibited the best performance:AUCtest=0.912,ACCtest=0.859,F1test=0.890.Moreover,it also demonstrated good predictive performance on the external validation set(AUCExt=0.944,RecallExt=0.856,FPRExt=0.111).This study provides a new technical route for KRAS inhibitor screening in natural product databases using artificial intelligence methods.

Objective:To evaluate the clinical value of cognitive and motor function in predicting conversion to neurodegenerative disorders in patients with isolated rapid eye movement sleep behavior disorder (iRBD).Methods:Forty-seven patients with iRBD were collected from the Department of Neurology of Tianjin Medical University General Hospital and Tianjin Medical University General Hospital Airport Site during October 2018 and June 2022. All participants received comprehensive evaluations of cognitive and motor function at baseline. The visuospatial function was evaluated by Rey-Osterrieth Complex Figure Test (ROCF)-copy, the memory function was evaluated by Auditory Verbal Learning Test and ROCF-recall, the attention-executive function was evaluated by Trail Making Test (TMT) and Stroop Color-Word Test, and the language function was evaluated by Boston Naming Test. The motor function was evaluated by Unified Parkinson′s Disease Rating Scale-Ⅲ, Alternate-tap Test (ATT), and 3-meter Timed Up and Go Test. The iRBD patients with phenoconversion were identified during follow-up. Receiver operating characteristic curve and generalized linear model Logistic regression were applied to identify the optimal combination of cognitive and motor tests in distinguishing the converters from non-converters in patients with iRBD. Multivariate Cox regression analyses were applied to evaluate the independent risk factors in predicting conversion to neurodegenerative diseases in patients with iRBD.Results:The median follow-up duration was 3 years. Forty-five iRBD patients were included in the analysis eventually, as 2 dropped out at follow-up. Twenty-one iRBD patients developed neurodegenerative disorders, with 14 presenting motor phenotype and 7 cognitive phenotype. Baseline ROCF-copy, TMT-A and ATT were best combination in identifying iRBD patients with phenoconversion [sensitivity: 90.0%, specificity: 87.5%, area under curve (AUC): 0.931, P<0.001]. Baseline TMT-A and ATT were best combination in identifying iRBD patients with motor phenotype conversion (sensitivity: 100.0%, specificity: 66.7%, AUC: 0.872, P<0.001); Baseline TMT-A performed best in identifying iRBD patients with cognitive phenotype conversion (sensitivity: 83.3%, specificity: 91.7%, AUC: 0.917, P<0.001). Multivariate Cox regression analysis showed that individuals with poorer performance of TMT-A (cut-off value: 63.0 s) and ATT (cut-off value: 205.5 taps/min) than the cut-off values at baseline had higher risks for developing to neurodegenerative disorders, with HR values of 5.455 (95% CI 1.243-23.941, P=0.025) and 11.279 (95% CI 1.485-85.646, P=0.019), respectively. Conclusions:In iRBD, ROCF-copy, TMT-A and ATT served as optimum combination in predicting phenoconversion, whereas TMT-A and ATT served as optimum combination in predicting motor phenotype, and TMT-A performed best in predicting cognitive phenotype. The performance in TMT-A and ATT in iRBD could predict the risk of developing to neurodegenerative disorders independently.