PFMBench: Protein Foundation Model Benchmark

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Gao, Zhangyang, Wang, Hao, Tan, Cheng, Xu, Chenrui, Liu, Mengdi, Hu, Bozhen, Chao, Linlin, Zhang, Xiaoming, Li, Stan Z.
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915348710686720
author Gao, Zhangyang
Wang, Hao
Tan, Cheng
Xu, Chenrui
Liu, Mengdi
Hu, Bozhen
Chao, Linlin
Zhang, Xiaoming
Li, Stan Z.
author_facet Gao, Zhangyang
Wang, Hao
Tan, Cheng
Xu, Chenrui
Liu, Mengdi
Hu, Bozhen
Chao, Linlin
Zhang, Xiaoming
Li, Stan Z.
contents This study investigates the current landscape and future directions of protein foundation model research. While recent advancements have transformed protein science and engineering, the field lacks a comprehensive benchmark for fair evaluation and in-depth understanding. Since ESM-1B, numerous protein foundation models have emerged, each with unique datasets and methodologies. However, evaluations often focus on limited tasks tailored to specific models, hindering insights into broader generalization and limitations. Specifically, researchers struggle to understand the relationships between tasks, assess how well current models perform across them, and determine the criteria in developing new foundation models. To fill this gap, we present PFMBench, a comprehensive benchmark evaluating protein foundation models across 38 tasks spanning 8 key areas of protein science. Through hundreds of experiments on 17 state-of-the-art models across 38 tasks, PFMBench reveals the inherent correlations between tasks, identifies top-performing models, and provides a streamlined evaluation protocol. Code is available at \href{https://github.com/biomap-research/PFMBench}{\textcolor{blue}{GitHub}}.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14796
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PFMBench: Protein Foundation Model Benchmark
Gao, Zhangyang
Wang, Hao
Tan, Cheng
Xu, Chenrui
Liu, Mengdi
Hu, Bozhen
Chao, Linlin
Zhang, Xiaoming
Li, Stan Z.
Biomolecules
Artificial Intelligence
Machine Learning
This study investigates the current landscape and future directions of protein foundation model research. While recent advancements have transformed protein science and engineering, the field lacks a comprehensive benchmark for fair evaluation and in-depth understanding. Since ESM-1B, numerous protein foundation models have emerged, each with unique datasets and methodologies. However, evaluations often focus on limited tasks tailored to specific models, hindering insights into broader generalization and limitations. Specifically, researchers struggle to understand the relationships between tasks, assess how well current models perform across them, and determine the criteria in developing new foundation models. To fill this gap, we present PFMBench, a comprehensive benchmark evaluating protein foundation models across 38 tasks spanning 8 key areas of protein science. Through hundreds of experiments on 17 state-of-the-art models across 38 tasks, PFMBench reveals the inherent correlations between tasks, identifies top-performing models, and provides a streamlined evaluation protocol. Code is available at \href{https://github.com/biomap-research/PFMBench}{\textcolor{blue}{GitHub}}.
title PFMBench: Protein Foundation Model Benchmark
topic Biomolecules
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2506.14796