arXiv 2506.067012025-06-07cs.LGcs.AIq-bio.BM

蛋白质 Transformer 是否具备生物智能？

Do Protein Transformers Have Biological Intelligence?

University of Delaware（德克萨斯大学）；Beijing University of Posts and Telecommunications（北京邮电大学）；Yale University（耶鲁大学）；University of Louisiana at Lafayette（路易斯安那州立大学）

Fudong Lin, Wanrou Du, Jinchan Liu, Tarikul Milon, Shelby Meche, Wu Xu, Xiaoqi Qin, Xu Yuan

备注：: Accepted by European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD 2025)

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文构建 Protein-FN 数据集，提出轻量高效的 SPT 架构与 Sequence Score 可解释方法，验证蛋白质 Transformer 能以高准确率发现符合生物学知识的序列模式。

摘要：

深度神经网络，尤其是 Transformer，已被广泛用于预测蛋白质的功能特性。在本研究中，我们聚焦于探究蛋白质 Transformer 是否能够捕捉蛋白质序列中的生物智能。为实现这一目标，我们首先引入一个蛋白质功能数据集，即 Protein-FN，它提供超过 9000 条具有有意义标签的蛋白质数据。其次，我们设计了一种新的 Transformer 架构，即 Sequence Protein Transformers（SPT），用于计算高效的蛋白质功能预测。第三，我们开发了一种名为 Sequence Score 的新型可解释人工智能（XAI）技术，该技术能够高效解释蛋白质模型的决策过程，从而克服解读隐藏在蛋白质 Transformer 中生物智能的困难。值得注意的是，即使是我们最小的 SPT-Tiny 模型（仅包含 5.4M 个参数）也展现出令人瞩目的预测准确率：在抗生素耐药性（AR）数据集上达到 94.3%，在 Protein-FN 数据集上达到 99.6%，且这些结果均通过从零开始训练取得。此外，我们的 Sequence Score 技术有助于揭示 SPT 模型能够发现蛋白质数据序列结构背后的若干有意义模式，这些模式与生物学界的领域知识高度吻合。我们已在 Hugging Face Datasets（https://huggingface.co/datasets/Protein-FN/Protein-FN）正式发布 Protein-FN 数据集。代码可在 https://github.com/fudong03/BioIntelligence 获取。

英文摘要：

Deep neural networks, particularly Transformers, have been widely adopted for predicting the functional properties of proteins. In this work, we focus on exploring whether Protein Transformers can capture biological intelligence among protein sequences. To achieve our goal, we first introduce a protein function dataset, namely Protein-FN, providing over 9000 protein data with meaningful labels. Second, we devise a new Transformer architecture, namely Sequence Protein Transformers (SPT), for computationally efficient protein function predictions. Third, we develop a novel Explainable Artificial Intelligence (XAI) technique called Sequence Score, which can efficiently interpret the decision-making processes of protein models, thereby overcoming the difficulty of deciphering biological intelligence bided in Protein Transformers. Remarkably, even our smallest SPT-Tiny model, which contains only 5.4M parameters, demonstrates impressive predictive accuracy, achieving 94.3% on the Antibiotic Resistance (AR) dataset and 99.6% on the Protein-FN dataset, all accomplished by training from scratch. Besides, our Sequence Score technique helps reveal that our SPT models can discover several meaningful patterns underlying the sequence structures of protein data, with these patterns aligning closely with the domain knowledge in the biology community. We have officially released our Protein-FN dataset on Hugging Face Datasets https://huggingface.co/datasets/Protein-FN/Protein-FN. Our code is available at https://github.com/fudong03/BioIntelligence.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。