Scholarlink

Abstract <p>We introduce <i>RES-LT</i> (Residual Local Topology), a novel topological data analysis (TDA) approach that extracts higher-order structural information from transformer-based protein language models. RES-LT utilizes both <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\({{H}_{0}}\)</EquationSource> <!--DANMath2570049Ivanova-m1--> </InlineEquation> and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({{H}_{1}}\)</EquationSource> <!--DANMath2570049Ivanova-m2--> </InlineEquation> persistent homology to characterize residue-residue interactions in proteins through attention maps, generating biologically relevant features for per-residue classification. Implemented on the ESM-2 model family, our framework integrates <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\({{H}_{0}}\)</EquationSource> <!--DANMath2570049Ivanova-m3--> </InlineEquation> and <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\({{H}_{1}}\)</EquationSource> <!--DANMath2570049Ivanova-m4--> </InlineEquation> topological features with standard embeddings to create a powerful hybrid representation. Extensive evaluation demonstrates that RES-LT achieves state-of-the-art performance in conservation prediction and significantly outperforms both traditional approaches and comparable transformer-based methods in binding site identification.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

RES-LT: Extracting Higher-Order Topological Features from Protein Language Models for Enhanced Per-Residue Classification

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月1日

查看全文
  • M. P. Ivanova,
  • I. E. Trofimov,
  • A. V. Mironenko,
  • P. V. Strashnov,
  • M. K. Kravchenko,
  • S. A. Barannikov,
  • E. V. Burnaev

摘要

Abstract

We introduce RES-LT (Residual Local Topology), a novel topological data analysis (TDA) approach that extracts higher-order structural information from transformer-based protein language models. RES-LT utilizes both \({{H}_{0}}\) and \({{H}_{1}}\) persistent homology to characterize residue-residue interactions in proteins through attention maps, generating biologically relevant features for per-residue classification. Implemented on the ESM-2 model family, our framework integrates \({{H}_{0}}\) and \({{H}_{1}}\) topological features with standard embeddings to create a powerful hybrid representation. Extensive evaluation demonstrates that RES-LT achieves state-of-the-art performance in conservation prediction and significantly outperforms both traditional approaches and comparable transformer-based methods in binding site identification.