Predicting protein electrostatics with protein language models
Abstract
Ionization states are crucial to protein function, yet predicting their equilibrium constants remains challenging. Here we present KaML-ESMs, sequence-based neural networks built on the ESM protein language models and trained on the GAINES-augmented PKAD-3r dataset. KaML-ESM2/ESMC outperform alternative approaches across four benchmarks, achieving root-mean-square errors of about 0.5 units for Asp/Glu/His/Cys/Tyr/Lys residues in native proteins. Performance degrades on 89 buried engineered mutations lacking evolutionary support, revealing a limitation that supervised training may help overcome. These results suggest that protein sequence, through its evolutionary context, encodes not only structure and function but also, indirectly, electrostatic characteristics. Applying KaML-ESM2 to the entire human proteome, we show that predicted p K a values can identify functional sites and infer catalytic mechanisms. We offer KaML, a sequence-based p K a prediction platform supporting biological exploration, protein engineering, and biomolecular simulation. With further development, GAINES could provide a general framework for addressing data scarcity in machine learning.
Related
- Predicting the evolutionary and functional landscapes of viruses with a unified nucleotide-protein language model: LucaVirus
- AmyloDeep: protein language model-based ensemble model for predicting amyloid propensity from the amino acid sequence
- Comparative cryo-EM structures of CGMMV and ToMMV reveal divergent coat protein packing and electrostatics indicating Tobamovirus host specificity
- An inverse-protein-folding pre-trained model for predicting the mutational effects in protein-protein interactions
- Understanding language model scaling for protein fitness prediction
BibTeX
@article{shen2026predicting,
title = {{Predicting protein electrostatics with protein language models}},
author = {Mingzhe Shen and Guy W. Dayhoff II and Daniel Kortzak and Jana Shen},
journal = {Nature Communications},
year = {2026},
month = oct,
doi = {10.1038/s41467-026-78406-y},
url = {https://www.nature.com/articles/s41467-026-78406-y},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predicting protein electrostatics with protein language models
AU - Mingzhe Shen
AU - Guy W. Dayhoff II
AU - Daniel Kortzak
AU - Jana Shen
PY - 2026
DA - 2026/10/24
JO - Nature Communications
DO - 10.1038/s41467-026-78406-y
UR - https://www.nature.com/articles/s41467-026-78406-y
AB - Ionization states are crucial to protein function, yet predicting their equilibrium constants remains challenging. Here we present KaML-ESMs, sequence-based neural networks built on the ESM protein language models and trained on the GAINES-augmented PKAD-3r dataset. KaML-ESM2/ESMC outperform alternative approaches across four benchmarks, achieving root-mean-square errors of about 0.5 units for Asp/Glu/His/Cys/Tyr/Lys residues in native proteins. Performance degrades on 89 buried engineered mutations lacking evolutionary support, revealing a limitation that supervised training may help overcome. These results suggest that protein sequence, through its evolutionary context, encodes not only structure and function but also, indirectly, electrostatic characteristics. Applying KaML-ESM2 to the entire human proteome, we show that predicted p K a values can identify functional sites and infer catalytic mechanisms. We offer KaML, a sequence-based p K a prediction platform supporting biological exploration, protein engineering, and biomolecular simulation. With further development, GAINES could provide a general framework for addressing data scarcity in machine learning.
SN - 2041-1723
ER -