Back

Accurate ΔTm Prediction Without Protein Structure Inputs for Biomolecular Stability

Siegismund, D.; Wieser, M.; Natali, E.; Steigele, S.

2026-07-06 bioinformatics
10.64898/2026.07.02.735991 bioRxiv
Show abstract

Predicting protein stability, like changes in melting temperature ({Delta}Tm) caused by mutations, is a critical task in therapeutic protein engineering and drug discovery. This is reflected by a growing solution space, including both AI-based sequence and structure based methods. This paper demonstrates that accurate {Delta}Tm prediction does not require structural input features, but can achieve state-of-the-art results with a careful training design for large sequence-based protein language models. We combine an autoresearch-inspired setup search with controlled ablation studies and show that a well-tuned sequence-only ESM2-650M model outperforms structure-informed methods in our benchmark, achieving the lowest error (MAE/RMSE) and competitive Pearson correlation without pH or structural inputs. We further show that choices such as loss function, pooling strategy, auxiliary supervision, and fine-tuning regime materially affect performance.

Matching journals

The top 7 journals account for 50% of the predicted probability mass.

1
Journal of Chemical Theory and Computation
140 papers in training set
Top 0.1%
15.0%
2
Journal of Chemical Information and Modeling
238 papers in training set
Top 0.7%
7.8%
3
Bioinformatics
1204 papers in training set
Top 3%
7.8%
4
Nature Machine Intelligence
70 papers in training set
Top 0.3%
7.2%
5
Briefings in Bioinformatics
354 papers in training set
Top 1%
6.7%
6
Computational and Structural Biotechnology Journal
242 papers in training set
Top 0.9%
4.3%
7
Nature Communications
5641 papers in training set
Top 32%
4.0%
50% of probability mass above
8
Nature Methods
385 papers in training set
Top 3%
3.2%
9
Protein Science
246 papers in training set
Top 1%
3.1%
10
PLOS Computational Biology
1863 papers in training set
Top 11%
2.6%
11
Bioinformatics Advances
203 papers in training set
Top 2%
2.6%
12
Scientific Reports
3612 papers in training set
Top 44%
2.4%
13
Cell Systems
201 papers in training set
Top 2%
2.4%
14
PRX Life
42 papers in training set
Top 0.4%
2.1%
15
Proceedings of the National Academy of Sciences
2444 papers in training set
Top 25%
2.0%
16
Journal of Cheminformatics
29 papers in training set
Top 0.4%
1.9%
17
Communications Biology
993 papers in training set
Top 15%
1.7%
18
Proteins: Structure, Function, and Bioinformatics
88 papers in training set
Top 0.9%
1.4%
19
eLife
5828 papers in training set
Top 54%
1.3%
20
Communications Chemistry
48 papers in training set
Top 0.8%
1.3%
21
Cell Reports Methods
165 papers in training set
Top 3%
1.1%
22
Journal of Molecular Biology
232 papers in training set
Top 3%
1.1%
23
Biophysical Journal
631 papers in training set
Top 4%
1.1%
24
Nucleic Acids Research
1281 papers in training set
Top 11%
1.1%
25
BMC Bioinformatics
457 papers in training set
Top 6%
0.8%
26
Nature Computational Science
55 papers in training set
Top 2%
0.8%
27
Human Genetics and Genomics Advances
84 papers in training set
Top 2%
0.8%