Attribution-based diagnostics for single-cell foundation models
Abstract
Single-cell RNA foundation models (scFMs) are increasingly used for in-silico biological discovery, yet predictive performance alone does not reveal whether their decisions rely on biologically meaningful signals or model-specific shortcuts. We systematically investigate post-hoc attribution of downstream predictions across two tasks: cell-type annotation using pretrained Cell2Sentence and scGPT models, and perturbation prediction using scGPT+STATE in a controlled setting. For cell annotation, we use attribution to assess whether predictions rely on established marker genes and how their relevance varies with model architecture, expression rank, and the surrounding cellular context. For perturbation prediction, we compare attribution patterns with known regulatory ground-truth in a synthetic setting to assess what they reveal about the model's learned computations. Across these tasks, attribution reveals biological feature use, contextual interactions, contributions from different input pathways, learning progression, associations with prediction quality, and potential pitfalls that are not fully captured by standard performance metrics. Experiments with complete and limited regulatory knowledge further demonstrate that conclusions drawn from attribution depend strongly on the choice of biological reference, baseline, and null model. Together, these findings highlight the value of attribution as a flexible post-hoc diagnostic for understanding model behavior, identifying potential issues, and informing the development of scFMs.
Related
- scRM: A single-cell foundation model trained on scRNA-seq reads
- Benchmarking single-cell foundation models reveals an advantage in the low-data regime for cell-type and disease-state prediction
- CellPatch: a flexible and efficient framework for single-cell foundation models empowered by heuristic gene patching
- DECONVersation: Single Cell Foundation Model-Derived Embeddings for Robust Cell Type Deconvolution of Bulk RNA-seq
- ProtCell: a foundation model for single-cell proteomics
BibTeX
@misc{jafari2026attributionbased,
title = {{Attribution-based diagnostics for single-cell foundation models}},
author = {Farnoush Rezaei Jafari and Tristan Lazard and Kenza Bouzid and Shruthi Bannur and Jeremiah Wander and Ashley Mae Conard and Stephanie Hyland and Lorin Crawford and Daniel Coelho de Castro},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.02335},
eprint = {2611.02335},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.02335},
}
RIS
TY - UNPB
TI - Attribution-based diagnostics for single-cell foundation models
AU - Farnoush Rezaei Jafari
AU - Tristan Lazard
AU - Kenza Bouzid
AU - Shruthi Bannur
AU - Jeremiah Wander
AU - Ashley Mae Conard
AU - Stephanie Hyland
AU - Lorin Crawford
AU - Daniel Coelho de Castro
PY - 2026
DA - 2026/11/02
PB - arXiv
DO - 10.48550/arxiv.2611.02335
UR - https://arxiv.org/abs/2611.02335
AB - Single-cell RNA foundation models (scFMs) are increasingly used for in-silico biological discovery, yet predictive performance alone does not reveal whether their decisions rely on biologically meaningful signals or model-specific shortcuts. We systematically investigate post-hoc attribution of downstream predictions across two tasks: cell-type annotation using pretrained Cell2Sentence and scGPT models, and perturbation prediction using scGPT+STATE in a controlled setting. For cell annotation, we use attribution to assess whether predictions rely on established marker genes and how their relevance varies with model architecture, expression rank, and the surrounding cellular context. For perturbation prediction, we compare attribution patterns with known regulatory ground-truth in a synthetic setting to assess what they reveal about the model's learned computations. Across these tasks, attribution reveals biological feature use, contextual interactions, contributions from different input pathways, learning progression, associations with prediction quality, and potential pitfalls that are not fully captured by standard performance metrics. Experiments with complete and limited regulatory knowledge further demonstrate that conclusions drawn from attribution depend strongly on the choice of biological reference, baseline, and null model. Together, these findings highlight the value of attribution as a flexible post-hoc diagnostic for understanding model behavior, identifying potential issues, and informing the development of scFMs.
ER -