MsPBRsP: Multi-scale Protein Binding Residues Prediction Using Language Model

Abstract

Accurate prediction of protein binding residues (PBRs) from sequence is important for the understanding of cellular activity and helpful for the design of novel drug. However, experimental methods are time-consuming and expensive. In recent years, a lot of computational predictors based on machine learning and deep learning models are proposed to reduce such consumption. But those methods often use MSA tools such as PSI-BLAST or NetSurfP to generate some statistical features and enter them into predictive models as necessary supplementary input. The input generation process normally takes long time, and there is no standard to specify which and how many statistic results should be provided to a prediction model. In addition, prediction of PBRs relies on residue local context, but the most appropriate scale is undetermined. Most works pre-selected certain residue features as input and a scale size based on expertise for certain type of PBRs. In this study, we propose a general tool-free end-to-end framework that can be applied to all types of PBRs, Multiscale Protein Binding Residues Prediction using language model (MsPBRsP). We adopt a pre-trained language model ProtTrans to save the large consumption caused by MSA tools, and use protein sequence alone as input to our model. To ease scale size uncertainty, we construct multi-size windows in attention layer and multi-size kernels in convolutional layer. We test our framework on various benchmark datasets including PBRs from protein-protein, protein-nucleotide, protein-small ligand, heterodimer, homodimer and antibody-antigen interactions. Compared with existing state-of-the-art methods, MsPBRsP achieves superior performance with less running time and higher prediction rates on every PBRs prediction task. Specifically, we boost F1 score by 27.1% and AUPRC score by 7.6% on NSP448 dataset and decrease running time from over 10 minutes to under 0.1s on average. The source code and datasets are available at https://github.com/biolushuai/MsPBRsP-for-multiple-PBRsprediction.

Large Language Models in the Clinic: A Comprehensive Benchmark

“…All three annotators have sufficient medical knowledge. In implementations, we follow previous works (Li et al, 2023b; Zhang et al, 2023b) to randomly select 200 real patient-doctor conversations from Li et al (2023b). We require the LLMs to simulate a doctor and provide responses based on various patient inquiries.…”

“…is an open-ended complex task and requires the models to first understand the real-world patient-clinician conversations, in which the conversation describes the conditions and symptoms, and then recommend all possible drugs for the treatment of patients. We use Chat-Doctor (Li et al, 2023b) for evaluation.…”

“…In detail, as shown in Table 3, ChatGLM-Med (Wang et al, 2023b) and DoctorGLM (Xiong et al, 2023) are fine-tuned on the ChatGLM-6B (Tsinghua KEG, 2023; Du et al, 2022; Zeng et al, 2022) using QA pairs and dialogues, respectively. Huatuo (Zhang et al, 2023a), ChatDoctor (Li et al, 2023b), Baize-Healthcare (Xu et al, 2023), and MedAlpaca-7B/13B (Han et al, 2023) are built upon the LLaMA-series models. During fine-tuning, both Huatuo and MedAlpaca employ the QA pairs collected from the medical knowledge graphs and medical texts, respectively.…”

“…Large language models (LLMs), such as ChatGPT (Brown et al, 2020; OpenAI, 2023b), LLaMA (Touvron et al, 2023a), and PaLM (Chowdhery et al, 2022), are increasingly being recognized for their potential in healthcare to aid clinical decision-making and provide innovative solutions for complex healthcare problems (Patel et al, 2023; Shen et al, 2023), e.g., discharge summary generation (Patel and Lam, 2023), health education (Safranek et al, 2023), and care planning (Fleming et al, 2023). Several recent efforts have been made to fine-tune publicly available general LLMs, e.g., LLaMA (Touvron et al, 2023b) and ChatGLM (Tsinghua KEG, 2023), to develop medical LLMs (Singhal et al, 2023a,c), resulting in ChatDoctor (Li et al, 2023b), MedAlpaca (Han et al, 2023), BenTsao (Wang et al, 2023a), and ClinicalCamel (Toma et al, 2023). Previous research shows that medical LLMs outperform human experts across a variety of medical tasks.…”

Starling: An I/O-Efficient Disk-Resident Graph Index Framework for High-Dimensional Vector Similarity Search on Data Segment

“…A notable advancement in this domain has been the widespread adoption of learning-based embedding models, which leverage high-dimensional vector representations to enable effective and efficient analysis and search of unstructured data [37,61]. High-dimensional Vector Similarity Search (HVSS) is a critical challenge in many domains, such as databases [25,68], information retrieval [28,32], recommendation systems [19,54], scientific computing [51,78], and large language models (LLMs) [7,12,44]. The computational complexity associated with exact query answering in HVSS has spurred recent research efforts toward developing approximate search methods [25,49,68].…”