Attribution assignment for deep-generative sequence models enables interpretability analysis using positive-only data
5 affiliations
- 1Department of Immunology and Oslo University Hospital, University of Oslo, Oslo, Norway
- 2Institute for Machine Learning, Johannes Kepler University Linz, Linz, Austria
- 3Department of Informatics, University of Oslo, Oslo, Norway
- 4Translational Immunology, Department of Biomedicine, University of Basel, Basel, Switzerland
- 5Imprint Labs, LLC, New York, New York, United States of America
Abstract
Generative machine learning models offer a powerful framework for therapeutic design, by efficiently exploring large spaces of biological sequences enriched for desirable properties. Unlike supervised learning methods, which require both positive and negative labeled data, generative models such as LSTMs can be trained solely on positively labeled sequences, for example, high-affinity antibodies. This is particularly advantageous in biological settings where negative data are scarce, unreliable, or biologically ill-defined. However, the lack of attribution methods for generative models has hindered the ability to extract interpretable biological insights from such models. To address this gap, we developed Generative Attribution Metric Analysis (GAMA), an attribution method for autoregressive generative models based on Integrated Gradients. We assessed GAMA using synthetic datasets with known ground truths to characterize its statistical behavior and validate its ability to recover biologically relevant features. We further demonstrated the utility of GAMA by applying it to experimental antibody-antigen binding data. GAMA enables model interpretability and the validation of generative sequence design strategies without the need for negative training data. Author summary: We have developed a new way to look inside AI language models that create biological sequences such as antibodies, even when only examples of successful binders are available. Traditional design methods need both positive and negative examples, but in many biomedical settings negative data are scarce, unreliable, or undefined. By training a model on only the desirable (positive) sequences and then comparing the model’s internal “gradient” signals to those of an untrained reference, we can identify which positions in a protein sequence the model deems important for binding. This approach, which we call Generative Attribution Metric Analysis (GAMA), lets us recover known binding motifs from synthetic data, confirm biologically relevant positions in simulated antibody‑antigen interactions, and highlight key residues in real experimental antibody datasets. The method therefore adds a layer of interpretability to otherwise “black-box” AI-based protein design tools, helping researchers understand why a generated sequence is predicted to work and enabling more confident use of these models in therapeutic development and other areas where only positive examples exist.
Related
- Generative AI for controllable protein sequence design: A survey
- Comparative Conformational Sampling of Apo Adenylate Kinase by Enhanced Molecular Dynamics and a Sequence-Based Generative Ensemble Model
- A Generative Neuro‐Symbolic AI for Protein Sequence Design
- Separating Detection Performance from Thermodynamic Interpretability in Quantum-Capacitance Affinity Assays.
- Fast and accurate taxonomic domain assignment of short metagenomic reads using BBERT
BibTeX
@article{frank2026attribution,
title = {{Attribution assignment for deep-generative sequence models enables interpretability analysis using positive-only data}},
author = {Robert Frank and Michael Widrich and Rahmad Akbar and Günter Klambauer and Geir Kjetil Sandve and Philippe A. Robert and Victor Greiff},
journal = {PLOS Computational Biology},
year = {2026},
month = sep,
doi = {10.1371/journal.pcbi.1014805},
url = {https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014805},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Attribution assignment for deep-generative sequence models enables interpretability analysis using positive-only data
AU - Robert Frank
AU - Michael Widrich
AU - Rahmad Akbar
AU - Günter Klambauer
AU - Geir Kjetil Sandve
AU - Philippe A. Robert
AU - Victor Greiff
PY - 2026
DA - 2026/09/28
JO - PLOS Computational Biology
DO - 10.1371/journal.pcbi.1014805
UR - https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1014805
AB - Generative machine learning models offer a powerful framework for therapeutic design, by efficiently exploring large spaces of biological sequences enriched for desirable properties. Unlike supervised learning methods, which require both positive and negative labeled data, generative models such as LSTMs can be trained solely on positively labeled sequences, for example, high-affinity antibodies. This is particularly advantageous in biological settings where negative data are scarce, unreliable, or biologically ill-defined. However, the lack of attribution methods for generative models has hindered the ability to extract interpretable biological insights from such models. To address this gap, we developed Generative Attribution Metric Analysis (GAMA), an attribution method for autoregressive generative models based on Integrated Gradients. We assessed GAMA using synthetic datasets with known ground truths to characterize its statistical behavior and validate its ability to recover biologically relevant features. We further demonstrated the utility of GAMA by applying it to experimental antibody-antigen binding data. GAMA enables model interpretability and the validation of generative sequence design strategies without the need for negative training data. Author summary: We have developed a new way to look inside AI language models that create biological sequences such as antibodies, even when only examples of successful binders are available. Traditional design methods need both positive and negative examples, but in many biomedical settings negative data are scarce, unreliable, or undefined. By training a model on only the desirable (positive) sequences and then comparing the model’s internal “gradient” signals to those of an untrained reference, we can identify which positions in a protein sequence the model deems important for binding. This approach, which we call Generative Attribution Metric Analysis (GAMA), lets us recover known binding motifs from synthetic data, confirm biologically relevant positions in simulated antibody‑antigen interactions, and highlight key residues in real experimental antibody datasets. The method therefore adds a layer of interpretability to otherwise “black-box” AI-based protein design tools, helping researchers understand why a generated sequence is predicted to work and enabling more confident use of these models in therapeutic development and other areas where only positive examples exist.
SN - 1553-7358
ER -