Personal academic homepage of Jiacai Yi — postdoctoral researcher in AI-enabled drug design and biomedical AI platforms at Hong Kong Baptist University.https://jiacai0101.github.io/https://jiacai0101.github.io/publications/commchem2025-drug-repositioning/https://jiacai0101.github.io/publications/commchem2025-drug-repositioning/This work presents a fused deep learning framework for drug repositioning. By integrating heterogeneous biomedical signals within a unified predictive architecture, the study improves prioritization of repurposing…Sat, 01 Nov 2025 00:00:00 GMThttps://jiacai0101.github.io/publications/chemsci2025-docking-limits/https://jiacai0101.github.io/publications/chemsci2025-docking-limits/Structure-based molecular docking, a cornerstone of computational drug design, is undergoing a paradigm shift fueled by deep learning (DL) innovations. However, the rapid proliferation of DL-driven docking methods…Wed, 01 Oct 2025 00:00:00 GMThttps://jiacai0101.github.io/publications/bib2025-fewshot-bayesian/https://jiacai0101.github.io/publications/bib2025-fewshot-bayesian/Hunting for candidate compounds with favorable pharmacological, toxicological, and pharmacokinetic properties in drug discovery is essentially a low-data problem, as data acquisition is both challenging and expensive.…Tue, 01 Jul 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250401/https://jiacai0101.github.io/posts/ai-drug-snapshot-250401/**期刊:** bioRxivTue, 01 Apr 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250402/https://jiacai0101.github.io/posts/ai-drug-snapshot-250402/**期刊**: arxivTue, 01 Apr 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250331/https://jiacai0101.github.io/posts/ai-drug-snapshot-250331/**期刊:** bioRxivMon, 31 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250330/https://jiacai0101.github.io/posts/ai-drug-snapshot-250330/**期刊:** bioRxivSun, 30 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250329/https://jiacai0101.github.io/posts/ai-drug-snapshot-250329/**期刊:** bioRxivSat, 29 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250328/https://jiacai0101.github.io/posts/ai-drug-snapshot-250328/**期刊:** bioRxivFri, 28 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250327/https://jiacai0101.github.io/posts/ai-drug-snapshot-250327/**期刊:** arxivThu, 27 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250326/https://jiacai0101.github.io/posts/ai-drug-snapshot-250326/**期刊:** bioRxivWed, 26 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250325/https://jiacai0101.github.io/posts/ai-drug-snapshot-250325/**期刊**: Computational Biology and ChemistryTue, 25 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250324/https://jiacai0101.github.io/posts/ai-drug-snapshot-250324/**期刊**: bioRxivMon, 24 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250323/https://jiacai0101.github.io/posts/ai-drug-snapshot-250323/**期刊:** NatureSun, 23 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250322/https://jiacai0101.github.io/posts/ai-drug-snapshot-250322/本文提出了名为AMCGRL的多领域协同图表示学习框架，用于分子相互作用预测，创新点在于借助跨域解码器和层次化互注意力机制捕捉复杂分子间相互作用模式。方法上，结合图卷积网络和图变换器，从域内和域间图中学习分子表示。实验在PepPI3966、MLI3099和RPI7317等数据集上开展，结果显示AMCGRL在分子表示学习能力上优于现有方法。总结表明，AMCGRL在分子相互作用预测任务中表现出色，...Sat, 22 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250321/https://jiacai0101.github.io/posts/ai-drug-snapshot-250321/本文提出了一种基于片段的蛋白质表示方法，通过使用40个进化保守的片段来显著降低蛋白质表示的维度，同时保留功能信息，创新地解决了传统蛋白质设计方法计算成本高的问题。该方法通过将蛋白质结构分解为片段图或片段集，利用滑动窗口算法和多种距离度量进行片段检测，并在PDBench和PFD数据集上验证了其有效性。实验结果表明，片段表示在功能聚类、数据库搜索和蛋白质设计中均优于传统方法，特别是在功能聚类中，...Fri, 21 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250320/https://jiacai0101.github.io/posts/ai-drug-snapshot-250320/本文提出了ABCFold，简化了AlphaFold 3、Boltz-1和Chai-1的运行与结果比较，通过标准化输入和统一输出格式，提升了用户体验和结果的可比性。该方法通过单一JSON输入文件自动转换为各工具所需格式，支持自定义多序列比对(MSA)和模板，并利用MMseqs2 API生成MSA，避免了本地数据库的下载需求。实验中使用UniRef30和colabfold_envdb数据集进行M...Thu, 20 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250319/https://jiacai0101.github.io/posts/ai-drug-snapshot-250319/本文提出了名为LigUnity的基础模型，通过联合优化虚拟筛选和先导化合物优化来预测蛋白质-配体结合亲和力，创新点在于将两个任务统一在一个模型中提升性能。方法上，利用对比学习进行虚拟筛选，采用列表排序方法进行先导化合物优化，并在八个基准数据集上验证。结果显示在DUD-E和DEKOIS 2.0基准上性能提升超50%，在未见蛋白质上泛化能力良好。实验表明，LigUnity在虚拟筛选和先导化合物优...Wed, 19 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250318/https://jiacai0101.github.io/posts/ai-drug-snapshot-250318/本文提出了DTIAM框架，用于预测药物-靶点相互作用、结合亲和力及药物作用机制，创新性地结合了自监督预训练和多任务学习。DTIAM通过从大量无标签数据中学习药物和靶点的表示，提高了下游任务的预测性能。实验在Yamanishi_08和Hetionet等数据集上进行，结果表明DTIAM在冷启动场景下表现优异，尤其在预测新药物或靶点时具有强大的泛化能力，验证了其在实际应用中的有效性。Tue, 18 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250316/https://jiacai0101.github.io/posts/ai-drug-snapshot-250316/本文提出了一种基于深度学习的策略，用于高效筛选GluN1/GluN3A受体抑制剂，创新点在于结合序列和复合物评分函数，平衡了筛选效率和准确性。方法上，首先使用序列评分函数从1800万化合物库中筛选出20,000个候选分子，随后采用IGModel和RTMScore两种复合物评分函数进行精确评分，最终通过全细胞电压钳电生理实验验证了活性分子PAT-505，其IC50为2.87±0.80μM。实验...Sun, 16 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250315/https://jiacai0101.github.io/posts/ai-drug-snapshot-250315/本文提出了一种名为DockEM的蛋白质-配体对接方法，创新性地利用中低分辨率冷冻电镜密度图进行精确对接。该方法通过局部密度图和物理能量优化，结合REMC模拟，实现了高精度配体定位。实验在DUDE和COACH数据集上进行了测试，结果显示DockEM在121个蛋白质-配体目标上表现优于其他先进对接方法。该方法显著提升了冷冻电镜密度图在虚拟药物筛选中的应用，为药物发现提供了更可靠的框架。Sat, 15 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250314/https://jiacai0101.github.io/posts/ai-drug-snapshot-250314/本文提出了一种名为DyRAMO的动态可靠性调整框架，用于在多目标分子设计中避免奖励黑客问题，其创新点在于通过动态调整预测模型的可靠性水平，确保设计的分子在多个目标属性上具有高预测可靠性。方法上，DyRAMO结合了贝叶斯优化和生成模型，通过迭代设置可靠性水平、设计分子和评估结果，实现了多目标优化与预测可靠性的平衡。实验以抗癌药物EGFR抑制剂设计为例，使用了ChemTSv2生成模型和多个预测模...Fri, 14 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250313/https://jiacai0101.github.io/posts/ai-drug-snapshot-250313/本文提出了一种基于折叠的端到端药物设计框架Llama-Gram，通过整合蛋白质折叠嵌入、图分子表示和不确定性估计，解决了传统大语言模型在药物设计中的幻觉问题。该方法利用冻结梯度的ESMFold模型和Graph Transformer变体，结合分组查询注意力机制和支持点理论的Gram层，提高了蛋白质-配体相互作用的预测准确性和可靠性。实验在ChEMBL 23、激酶和GPCR数据集上进行，Lla...Thu, 13 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250312/https://jiacai0101.github.io/posts/ai-drug-snapshot-250312/本文提出了一种名为CheapVS的新型框架，通过结合偏好多目标贝叶斯优化和扩散对接模型，显著提高了药物发现中虚拟筛选的效率和可靠性。该方法利用化学家的成对偏好反馈，优化多个药物属性（如结合亲和力、溶解度和毒性），从而在有限的计算预算内识别出更多潜在药物。实验在包含10万个化学候选物的EGFR靶点库上进行，结果显示CheapVS在仅扫描6%的库时能够恢复37种已知药物中的16种，优于现有对接方...Wed, 12 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250311/https://jiacai0101.github.io/posts/ai-drug-snapshot-250311/本文提出了一种名为KGDRP的知识引导图学习方法，通过整合多模态生物医学数据，显著提升了表型筛选和靶点发现的预测性能。该方法利用异质图神经网络（HGNN）构建生物医学异质图（BioHG），结合基因表达数据、化学分子结构和生物网络信息，解决了药物冷启动问题，并在药物反应预测和靶点发现任务中表现出色。实验使用了GDSC数据集进行药物反应预测，并在COVID-19零样本评估中展示了较高的成功率，结...Tue, 11 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250310/https://jiacai0101.github.io/posts/ai-drug-snapshot-250310/本文提出了一种基于多模态数据融合的图对比学习框架（MDFCL），用于分子性质预测，创新点在于结合了分子图和序列结构，设计了自适应数据增强策略，并通过对比学习框架进行预训练。方法上，MDFCL通过图神经网络和卷积网络分别编码分子图和序列数据，利用自适应增强策略生成四种增强样本，并通过对比学习优化模型。实验在13个分子性质预测基准数据集上进行，包括BBBP、Tox21、ClinTox等，结果表明...Mon, 10 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/posts/ai-drug-snapshot-250309/https://jiacai0101.github.io/posts/ai-drug-snapshot-250309/本文提出了一种基于SE(3)流匹配的蛋白质构象集合生成模型P2DFlow，通过引入ESMFold预测加扰动的强先验和近似能量机制，显著提升了蛋白质构象集合的预测精度和多样性。该方法利用SE(3)等变流匹配框架，结合ESMFold预测和扰动生成先验分布，并通过引入近似能量维度来区分不同构象，从而有效避免生成不存在的中间态。实验在ATLAS数据集上进行，P2DFlow在有效性、保真度和动力学指标...Sun, 09 Mar 2025 00:00:00 GMThttps://jiacai0101.github.io/publications/nar2024-ddinter2/https://jiacai0101.github.io/publications/nar2024-ddinter2/Drug interactions pose significant challenges in clinical practice, potentially leading to adverse drug reactions, reduced efficacy, and even life-threatening consequences. DDInter 2.0 substantially expands data…Wed, 01 Jan 2025 00:00:00 GMThttps://jiacai0101.github.io/publications/nar2024-admetlab3/https://jiacai0101.github.io/publications/nar2024-admetlab3/ADMETlab 3.0 is the second updated version of the web server that provides a comprehensive and efficient platform for evaluating ADMET-related parameters as well as physicochemical properties and medicinal chemistry…Mon, 01 Jul 2024 00:00:00 GMThttps://jiacai0101.github.io/publications/nar2024-chemfh/https://jiacai0101.github.io/publications/nar2024-chemfh/High-throughput screening rapidly tests extensive arrays of chemical compounds to identify hit compounds for specific biological targets in drug discovery. However, false-positive results disrupt hit-to-lead progression…Mon, 01 Jul 2024 00:00:00 GMThttps://jiacai0101.github.io/publications/database2024-istransbase/https://jiacai0101.github.io/publications/database2024-istransbase/ISTransbase is an online knowledgebase for drug transporters that systematically collects inhibitor and substrate data from the literature. It supports transporter-centered exploration of drug transport mechanisms and…Sat, 01 Jun 2024 00:00:00 GMThttps://jiacai0101.github.io/publications/natprotoc2024-optadmet/https://jiacai0101.github.io/publications/natprotoc2024-optadmet/Lead optimization is a crucial step in drug discovery, aiming to design potential drug candidates from biologically active hits. During lead optimization, active hits undergo modifications to achieve a balance between…Mon, 01 Apr 2024 00:00:00 GMThttps://jiacai0101.github.io/publications/bib2024-chemmort/https://jiacai0101.github.io/publications/bib2024-chemmort/Drug discovery and development constitute a laborious and costly undertaking. The success of a drug hinges not only on good efficacy but also on acceptable ADMET properties. ChemMORT is an automatic ADMET optimization…Fri, 01 Mar 2024 00:00:00 GMThttps://jiacai0101.github.io/publications/see2023-space-cyber-ontology/https://jiacai0101.github.io/publications/see2023-space-cyber-ontology/This paper develops an ontology modeling framework for threat awareness in space cyber operations. The resulting ontology supports structured representation, reasoning, and analysis of satellite cyberspace threat…Wed, 01 Mar 2023 00:00:00 GMThttps://jiacai0101.github.io/publications/bioinfo2022-micer/https://jiacai0101.github.io/publications/bioinfo2022-micer/Automatic recognition of chemical structures from molecular images provides an important avenue for the rediscovery of chemicals. Traditional rule-based approaches rely on expert knowledge and struggle with diverse…Sat, 01 Oct 2022 00:00:00 GMThttps://jiacai0101.github.io/publications/see2022-satellite-cyber-situational/https://jiacai0101.github.io/publications/see2022-satellite-cyber-situational/This work studies satellite cyberspace situational understanding through ontology-guided knowledge reasoning. It builds reasoning rules and a domain knowledge base to support automatic correlation analysis of…Sun, 01 May 2022 00:00:00 GMThttps://jiacai0101.github.io/publications/bib2022-abc-net/https://jiacai0101.github.io/publications/bib2022-abc-net/ABC-Net proposes a divide-and-conquer architecture for translating molecular images into SMILES strings. By decomposing the recognition problem into coordinated sub-tasks, the model improves robustness across diverse…Tue, 01 Mar 2022 00:00:00 GMThttps://jiacai0101.github.io/publications/nar2022-ddinter/https://jiacai0101.github.io/publications/nar2022-ddinter/Drug-drug interaction (DDI) can trigger many adverse effects in patients and has emerged as a threat to medicine and public health. We present DDInter, a curated DDI database with comprehensive data, practical…Sat, 01 Jan 2022 00:00:00 GMThttps://jiacai0101.github.io/publications/research2022-smiles-enumeration-bert/https://jiacai0101.github.io/publications/research2022-smiles-enumeration-bert/This work explores multitask molecular property prediction with a BERT framework enhanced by SMILES enumeration. The study shows that large-scale pretraining and sequence augmentation can improve robustness and…Sat, 01 Jan 2022 00:00:00 GMThttps://jiacai0101.github.io/publications/bmcbioinfo2021-microbe-disease/https://jiacai0101.github.io/publications/bmcbioinfo2021-microbe-disease/This paper presents a transfer learning framework for extracting microbe-disease interactions from biomedical literature. The model improves relation mining performance under limited labeled data and supports automated…Wed, 01 Sep 2021 00:00:00 GMThttps://jiacai0101.github.io/publications/nar2022-admetlab2/https://jiacai0101.github.io/publications/nar2022-admetlab2/Because undesirable pharmacokinetics and toxicity of candidate compounds are the main reasons for the failure of drug development, it has been widely recognized that ADMET should be evaluated as early as possible. Here,…Thu, 01 Jul 2021 00:00:00 GMThttps://jiacai0101.github.io/publications/bib2021-mg-bert/https://jiacai0101.github.io/publications/bib2021-mg-bert/MG-BERT introduces a BERT-style framework for unsupervised atomic representation learning from molecular structures. The learned representations improve downstream molecular property prediction and demonstrate the value…Fri, 01 Jan 2021 00:00:00 GMT