研究方向
1）AI底層技術(shù)：通過在模型中注入知識/規(guī)則，提升分子性質(zhì)預測模型的泛化能力；
2）智能化學合成：利用深度學習、專家規(guī)則、量子化學構(gòu)建化學合成智能體；
3）智能生物合成：利用AI設計、優(yōu)化萜類合成酶和P450酶，輔助類藥分子合成；
4）蛋白詞表技術(shù)：從大模型中提取具有功能的蛋白語義單元（蛋白質(zhì)的“官能團”），輔助藥物、酶的設計。
實驗室主頁：https://www.tianlab-tsinghua.com/
學術(shù)貢獻
田博學實驗室近5年在化學信息學與AI制藥領(lǐng)域取得系統(tǒng)性的研究成果，以通訊作者（含共同）身份發(fā)表系列研究論文，具體總結(jié)如下：
1）發(fā)展了蛋白詞表技術(shù)，利用無監(jiān)督學習分析大模型的注意力矩陣，解析蛋白語義單元，并利用監(jiān)督學習映射到功能，有望成為下一代蛋白模擬、蛋白設計、藥物設計、細胞模擬技術(shù)的基礎(chǔ) (B.Tian* BioRxiv 2025)。
2）發(fā)展了物理啟發(fā)式蛋白語言模型，利用遠端突變提升了萜類合成酶的產(chǎn)物產(chǎn)量（70倍以上），輔助類藥分子合成 (J. Qiao* B.Tian* Research Square 2024)。
3）發(fā)展了基于蛋白質(zhì)語言模型的蛋白-DNA、蛋白-小分子的結(jié)合位點預測方法（B.Tian* Breif. Bioinform. 2024，B.Tian* J. Cheminform. 2024），抗體CDR結(jié)構(gòu)預測 (F. Qian*, B.Tian* eLife 2024)、P450底物、產(chǎn)物預測方法 (B.Tian* J. Chem. Inf. Model. 2024，B.Tian* J. Chem. Inf. Model. 2025)，輔助藥物發(fā)現(xiàn)。
4）揭示了深度學習方法在分子性質(zhì)預測中泛化能力不足的問題，訓練獲得了在MoleculeNet和MolData數(shù)據(jù)集上當前最佳模型（B.Tian* J. Chem. Inf. Model. 2025， B.Tian* J. Chem. Theory Comput. 2024），為化學信息學和AI制藥領(lǐng)域提供了新思路。
5）在膜蛋白生物學機制研究方面，開展了轉(zhuǎn)運蛋白(B.Tian, Y. Yuan*, C. Yan* Nature 2024)、離子通道(B.Tian*, X.Li*, B. Xiao* Neuron 2025)等體系的計算模擬。
6）利用量子化學方法研究了復雜環(huán)化、偶聯(lián)、重排反應機制（B.Tian*, W.Deng* ACS Catal. 2021，W. Yang*，B.Tian*, W.Deng* ACS Catal. 2021， B.Tian*, X.Liao* Cell Rep. Phys. Sci. 2023， B.Tian*, Y.Tang* Org. Chem. Front. 2022）。
近5年論文發(fā)表情況：
1. Automatically Defining Protein Words for Diverse Functional Predictions Based on Attention Analysis of a Protein Language Model. Hedi Chen, Jingrui Zhong, Xiaochun Zhang, Jingke Chen, Lin Guo, Xiaoliang Xiong, Xiaonan Zhang, Xiangyu Liu, Bailong Xiao, Boxue Tian* bioRxiv 2025, https://doi.org/10.1101/2025.01.20.633699
2. GLMCyp: A Deep Learning-Based Method for CYP450-Mediated Reaction Site Prediction. Xuhai Huang, Jiamin Chang, Boxue Tian* Journal of Chemical Information and Modeling 2025, 65(5):2322-2335
3. SynthMol: A Drug Safety Prediction Framework Integrating Graph Attention and Molecular Descriptors into Pre-Trained Geometric Models. Zidong Su, Rong Zhang, Xiaoyu Fan, Boxue Tian* Journal of Chemical Information and Modeling 2025, 65：2256?2267
4. An intermediate open structure reveals the gating transition of the mechanically activated PIEZO1 channel. Sijia Liu ,Xuzhong Yang, Xudong Chen, Xiaochun Zhang, Jinghui Jiang, Jingyi Yuan, Wenhao Liu, Li Wang, Heng Zhou, Kun Wu, Boxue Tian*, Xueming Li*, Bailong Xiao* Neuron 2025,113,1-15
5. Enhancing Enzyme Activity with Mutation Combinations Guided by Few-shot Learning and Causal Inference. Research Square 2024, https://www.researchsquare.com/article/rs-5354708/v1
6. Protein–DNA binding sites prediction based on pre-trained protein language model and contrastive learning. Yufan Liu, Boxue Tian* Briefings in Bioinformatics 2024, 25(1):1-12
7. DeepP450: Predicting human P450 activities of small molecules by integrating pretrained protein language model and molecular representation. Jiamin Chang, Xiaoyu Fan, Boxue Tian* Journal of Chemical Information and Modeling 2024, 64(18):3149-3160
8. PointGAT: A quantum chemical property prediction model integrating graph attention and 3D geometry. Rong Zhang, Rongqing Yuan, Boxue Tian* Journal of Chemical Theory and Computation 2024, 20(10):4115-4128
9. Accurate prediction of CDR-H3 loop structures of antibodies with deep learning Hedi Chen,Xiaoyu Fan,Shuqian Zhu,Yuchan Pei,Xiaochun Zhang,Xiaonan Zhang,Lihang Liu,Feng Qian*,Boxue Tian* eLife 2024 12:RP91512
10. Molecular basis of human noradrenaline transporter reuptake and inhibition Jiaxin Tan,Yuan Xiao,Fang Kong,Xiaochun Zhang,Hanwen Xu,Angqi Zhu,Yiming Liu,Jianlin Lei,Boxue Tian,Yafei Yuan*,Chuangye Yan* Nature 2024, 632:921-929
11. Protein-small molecule binding site prediction based on a pre-trained protein language model with contrastive learning. Jue Wang, Yufan Liu & Boxue Tian* Journal of Cheminformatics 2024 16(1):125
12. Coupling of alkynes and aryl halides with nickel-catalyzed Sonogashira reactions. Hui Chen,Licheng Yao,Lin Guo,Yahu A. Liu,Boxue Tian,Xuebin Liao* Cell Reports Physical Science 2023, 5(1):102900
13. Elucidating the selectivity of dyotropic rearrangements of β-lactones: a computational survey. Jingyang Zhang, Yumiao Ma, Ke Qiu, Bo Li, Zhengwen Xue, Boxue Tian*,Yefeng Tang* Organic Chemistry Frontiers 2022, 9:329-341
14. Nickel-catalyzed enantioselective domino Heck/Sonogashira coupling for construction of C(sp)-C(sp3) bond-substituted quaternary carbon centers. Hui Chen, Zhenkang Ai, Lingmei Guo,Licheng Yao, Yaopeng Li, Buming Gu, Yixuan Zhang, Yahu A. Liu, Boxue Tian*, Xuebin Liao* Tetrahedron Chem 2022, 2:100021
15. Cooperative N-heterocyclic Carbene and Iridium Catalysis Enables Stereoselective and Regiodivergent [3 + 2] and [3 + 3] Annulation Reactions. Jian Zhang, Yanshan Gao, Buming Gu, Wulin Yang*, Boxue Tian*,Weiping Deng* ACS Catalysis 2021 11(7):3810-3821
16. Diastereo- and Enantioselective Synthesis of Eight-Membered Heterocycles via an Allylation/Ring Expansion Sequence Enabled by Multiple Catalysis. Wulin Yang*,Yuanlin Wang, Wen Li, Buming Gu, Siwen Wang, Xiaoyan Luo, Boxue Tian*, Weiping Deng* ACS Catalysis 2021 11(20):12557-12564
