PDFBench: A Benchmark for De Novo Protein Design from Function

Jiahao Kuang, Nuowei Liu, Changzhi Sun, Jie Wang, Tao Ji, Yuanbin Wu
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:61349-61371, 2026.

Abstract

Function-guided protein design is a crucial task with significant applications in drug discovery and enzyme engineering. However, the field lacks a unified and comprehensive evaluation framework. Current models are assessed using inconsistent and limited subsets of metrics, which prevents fair comparison and a clear understanding of the relationships between different evaluation criteria. To address this gap, we introduce PDFBench, the first comprehensive benchmark for function-guided de novo protein design. Our benchmark systematically evaluates eight state-of-the-art models on 16 metrics across two key settings: description-guided design, for which we repurpose the Mol-Instructions dataset, originally lacking quantitative benchmarking, and keyword-guided design, for which we introduce a new test set, SwissTest, created with a strict datetime cutoff to ensure data integrity. By benchmarking across a wide array of metrics and analyzing their correlations, PDFBench enables more reliable model comparisons and provides key insights to guide future research.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-kuang26d, title = {{PDFB}ench: A Benchmark for De Novo Protein Design from Function}, author = {Kuang, Jiahao and Liu, Nuowei and Sun, Changzhi and Wang, Jie and Ji, Tao and Wu, Yuanbin}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {61349--61371}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/kuang26d/kuang26d.pdf}, url = {https://proceedings.mlr.press/v306/kuang26d.html}, abstract = {Function-guided protein design is a crucial task with significant applications in drug discovery and enzyme engineering. However, the field lacks a unified and comprehensive evaluation framework. Current models are assessed using inconsistent and limited subsets of metrics, which prevents fair comparison and a clear understanding of the relationships between different evaluation criteria. To address this gap, we introduce PDFBench, the first comprehensive benchmark for function-guided de novo protein design. Our benchmark systematically evaluates eight state-of-the-art models on 16 metrics across two key settings: description-guided design, for which we repurpose the Mol-Instructions dataset, originally lacking quantitative benchmarking, and keyword-guided design, for which we introduce a new test set, SwissTest, created with a strict datetime cutoff to ensure data integrity. By benchmarking across a wide array of metrics and analyzing their correlations, PDFBench enables more reliable model comparisons and provides key insights to guide future research.} }
Endnote
%0 Conference Paper %T PDFBench: A Benchmark for De Novo Protein Design from Function %A Jiahao Kuang %A Nuowei Liu %A Changzhi Sun %A Jie Wang %A Tao Ji %A Yuanbin Wu %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-kuang26d %I PMLR %P 61349--61371 %U https://proceedings.mlr.press/v306/kuang26d.html %V 306 %X Function-guided protein design is a crucial task with significant applications in drug discovery and enzyme engineering. However, the field lacks a unified and comprehensive evaluation framework. Current models are assessed using inconsistent and limited subsets of metrics, which prevents fair comparison and a clear understanding of the relationships between different evaluation criteria. To address this gap, we introduce PDFBench, the first comprehensive benchmark for function-guided de novo protein design. Our benchmark systematically evaluates eight state-of-the-art models on 16 metrics across two key settings: description-guided design, for which we repurpose the Mol-Instructions dataset, originally lacking quantitative benchmarking, and keyword-guided design, for which we introduce a new test set, SwissTest, created with a strict datetime cutoff to ensure data integrity. By benchmarking across a wide array of metrics and analyzing their correlations, PDFBench enables more reliable model comparisons and provides key insights to guide future research.
APA
Kuang, J., Liu, N., Sun, C., Wang, J., Ji, T. & Wu, Y.. (2026). PDFBench: A Benchmark for De Novo Protein Design from Function. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:61349-61371 Available from https://proceedings.mlr.press/v306/kuang26d.html.

Related Material