De novo Protein Generating Model via Structure-Aware Graph Neural Networks
- 1University of Texas at Dallas
Abstract
The focus of this paper is to create a protein conformation prediction model that is optimized for structural accuracy rather than sequence recovery. To achieve this goal, we created a graph neural network (GNN) [19] for the design of computational proteins that integrates differentiable structure prediction through ESMFold [13] via a structural loss function, allowing direct optimization of the 3D conformational accuracy. Our model takes protein backbone coordinates as input and predicts an amino acid sequence using a composite loss that heavily weights structural similarity. On a matched evaluation sample, our model achieves structural precision (RMSD, TM-score, and GDT-TS) that is statistically indistinguishable from ProteinMPNN, a leading sequence recovery-focused method, despite training with no direct sequence recovery supervision and substantially lower native sequence recovery. Our results demonstrate that sequence recovery is not necessary to achieve competitive structural accuracy and that proteins can possess multiple non-unique solutions, suggesting that protein sequence prediction models may perform comparably well under purely structural conformation training. This distinction in training data and model structure allows for the training of a protein-generating model on non-natural proteins. Creating a model designed for de novo protein generation.
Related
- Protein Language Model-Conditioned Graph Neural Networks for Multitask GPCR Ligand Activity Prediction
- Relation-Aware Graph-Hypergraph Networks for Antibody Sequence-Structure Co-Design
- Covariate-aware genomic prediction of blood metabolite profiles using multi-task neural networks
- Representation-aware calibration enables statistically reliable protein structure search
- StableProt: Structure-Aware Deep Learning for Protein Thermostability (Tm) and Environmental Adaptation (OGT) Prediction
BibTeX
@misc{gammill2026novo,
title = {{De novo Protein Generating Model via Structure-Aware Graph Neural Networks}},
author = {C. M. Gammill and L. Tamil},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.27.762190},
url = {https://www.biorxiv.org/content/10.64898/2026.10.27.762190v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - De novo Protein Generating Model via Structure-Aware Graph Neural Networks
AU - C. M. Gammill
AU - L. Tamil
PY - 2026
DA - 2026/10/27
PB - bioRxiv
DO - 10.64898/2026.10.27.762190
UR - https://www.biorxiv.org/content/10.64898/2026.10.27.762190v1
AB - The focus of this paper is to create a protein conformation prediction model that is optimized for structural accuracy rather than sequence recovery. To achieve this goal, we created a graph neural network (GNN) [19] for the design of computational proteins that integrates differentiable structure prediction through ESMFold [13] via a structural loss function, allowing direct optimization of the 3D conformational accuracy. Our model takes protein backbone coordinates as input and predicts an amino acid sequence using a composite loss that heavily weights structural similarity. On a matched evaluation sample, our model achieves structural precision (RMSD, TM-score, and GDT-TS) that is statistically indistinguishable from ProteinMPNN, a leading sequence recovery-focused method, despite training with no direct sequence recovery supervision and substantially lower native sequence recovery. Our results demonstrate that sequence recovery is not necessary to achieve competitive structural accuracy and that proteins can possess multiple non-unique solutions, suggesting that protein sequence prediction models may perform comparably well under purely structural conformation training. This distinction in training data and model structure allows for the training of a protein-generating model on non-natural proteins. Creating a model designed for de novo protein generation.
ER -