陈文杰
Resume简历 Contact联系

About关于

Chen Wenjie · 陈文杰

Building interpretable AI systems for proteins, biological data, and computational discovery.为蛋白质和生物数据构建可解释的 AI 系统，让计算真正参与发现。

Position研究定位

Questions before models问题先于模型

I work on AI for biology: protein language models, computational protein design, biological representation learning, and trustworthy biomedical AI. The recurring question is simple to state and difficult to answer: what do these models actually know, and how do we prove it?我做生物 AI 方向的研究：蛋白质语言模型、计算蛋白质设计、生物表征学习和可信生物医学 AI。有一个问题反复出现，它很好说清，却很难回答：这些模型究竟知道什么？我们又怎么证明？

I am an undergraduate in Bioengineering at Wuhan Institute of Technology, with an expected graduation date of June 2027. My research path began with SERS/MOF materials experiments and has since moved toward bioinformatics, protein computation, and machine learning.我目前就读于武汉工程大学生物工程专业，预计于 2027 年 6 月本科毕业。我的研究路径始于 SERS/MOF 材料实验，随后逐步转向生物信息学、蛋白质计算与机器学习。

Selected work代表性工作

Research experience研究经历

  1. 2026–present Manuscript Submitted稿件已投稿

    SAE-DISTILLSAE-DISTILL

    Testing whether sequence-only models can recover selected SAE-derived residue profiles from ESM-C representations.检验仅使用序列的轻量模型能否恢复 ESM-C 表征中选定的 SAE 残基特征谱。

    First author — developed the study, implemented the computational workflow, analysed the results, and prepared the manuscript.第一作者——负责研究设计、计算流程实现、结果分析和稿件撰写。

  2. 2026–present Manuscript Submitted稿件已投稿

    PAO1 Anti-virulence DiscoveryPAO1 抗毒力候选物优先级排序

    Evidence-stratified computational prioritization of natural-product candidates across virulence-associated PAO1 targets.针对 PAO1 毒力相关靶点，对天然产物候选物进行分层证据支持的计算优先级排序。

    First author — developed the multi-criteria prioritization workflow, analysed the computational evidence, and prepared the manuscript.第一作者——构建多指标优先级排序流程，分析计算证据并完成稿件撰写。

  3. 2024–2026 Published research已发表研究

    SERS / MOF materials researchSERS / MOF 材料研究

    Experimental research on SERS/MOF composites for food-safety monitoring and small-molecule biomarker detection.参与面向食品安全检测和小分子生物标志物检测的 SERS/MOF 复合材料实验研究。

    Co-author — contributed to material preparation, sample testing, data organization, manuscript writing, and supporting electric-field simulations.共同作者——参与材料制备、样品检测、数据整理与论文写作，并辅助完成部分电场模拟。

Research practice研究实践

Working principles工作原则

  1. 01

    Ask a falsifiable question提出可证伪的问题

    Start from what would disprove the idea, not from which model is fashionable.先想清楚什么结果会否定这个想法，而不是先挑当下流行的模型。

  2. 02

    Make the split resemble use数据划分要贴近真实用法

    Family-held-out proteins and external clinical cohorts expose failure modes that random splits can hide.随机划分会掩盖一些失败模式，蛋白质家族留出和外部临床队列能把它们暴露出来。

  3. 03

    Stress-test the explanation解释也要经受压力测试

    Structured controls help separate a recovered biological signal from family texture or benchmark leakage.真正恢复的生物信号、家族统计纹理、基准泄漏，结构化对照能把这三者分开。

  4. 04

    Stop the claim at the evidence论断止于证据

    Computational prioritization, calibration and interpretation each support different kinds of sentences.计算优先排序、校准、解释，各自能撑起的话不一样。

Research map研究图谱

Research interests研究兴趣

  • AI for Biology生物人工智能
  • Computational Biology计算生物学
  • Computational Protein Design计算蛋白质设计
  • Protein Language Models蛋白质语言模型
  • Biological Representation Learning生物表征学习
  • Interpretable Machine Learning可解释机器学习
  • Bioinformatics生物信息学
  • AI-assisted Drug DiscoveryAI 辅助药物发现