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.

Objective:To evaluate the efficacy and safety of preoperative intravitreal conbercept (IVC) as an adjunct to pars plana vitrectomy (PPV) in the treatment of proliferative diabetic retinopathy (PDR).Methods:A systematic search in EMbase, PubMed, Cochrane Library, Chinese periodical full text database (CNKI), Wanfang database and VIP database were conducted, studies about the effectiveness and safety of IVC combined with PPV in the management of PDR were collected.Two researchers independently screened the studies according to the inclusion criteria and exclusion criteria, and extracted the data.The quality of the randomized controlled trial (RCT) was evaluated by the modified Jadad scale, and the quality of the cohort study or case-control study was evaluated by the Newcastle Ottawa scale (NOS). Rev Man 5.3 was applied for data analysis.Results:A total of 11 RCTs, 2 cohort studies and 10 case-control studies involving 1 625 patients and 1 844 eyes were included.The final Jadad score for each RCT was more than 3, and the final NOS score for each cohort study and case-control study was more than 5.The results of Meta-analysis showed that the total effectiveness of treatment was significantly higher in the preoperative IVC group than that in the simple PPV group ( RR=1.31, 95% CI: 1.21-1.42, P<0.001). The average operation duration was significantly shorter in the preoperative IVC group compared with that in the simple PPV group (MD=-21.11, 95% CI: -26.39--15.83, P<0.001). The level of VEGF was significantly lower in the preoperative IVC group than that in the simple PPV group (MD=-15.33, 95% CI: -19.40--11.26, P<0.001). Preoperative IVC could reduce the incidences of intraoperative bleeding, iatrogenic retinal breaks, postoperative recurrent vitreous hemorrhage and temporary increase of intraocular pressure, with statistically significant differences between them (all at P<0.05). Conclusions:Preoperative intravitreal injection of conbercept shows better effect and safer than vitrectomy alone, and it has no serious side effect.