Scooby-2: fine-tuning AlphaGenome for single-cell genomic track prediction
- 1Technical University of Munich
Abstract
Sequence-to-function models predict genomic tracks from DNA sequence. AlphaGenome has been shown to outperform previous models, including Borzoi, on most track-prediction and variant-effect benchmarks. However, like its predecessors, it predicts tracks only for predefined, mostly bulk-derived cell types and tissues. While Scooby has extended Borzoi to single-cell resolution, a single-cell adaptation of AlphaGenome is still lacking. We present Scooby-2, which combines AlphaGenome with a cell-specific decoder to predict single-cell RNA-seq coverage and chromatin-accessibility genomic tracks from DNA sequence. The package provides the implementation, the parameter-efficient fine-tuning setup, the training pipeline and inference scripts. When evaluated on the NeurIPS bone marrow single-cell dataset, Scooby-2 achieved a Pearson correlation of 0.56 when predicting cell-type-specific gene expression differences, compared with 0.53 for Scooby. When using the distilled AlphaGenome backbone, variant effect prediction concordance of fine-mapped eQTLs from the OneK1K cohort improved from 0.37 to 0.47 Spearman's rank correlation across all associations, an improvement consistent with that of AlphaGenome over Borzoi itself.
Related
- AlphaGenome informed machine learning for pathogenicity prediction of germline mutations
- Genomic prediction of single cross families of perennial ryegrass under two nitrogen managements
- Analysis of single nucleotide substitution in CYP21A2 gene using Alphagenome
- Response representation shapes AlphaGenome enhancer-gene prediction in human astrocytes
- Beyond a single threshold: calibration of CADD v1.7 and AlphaGenome for ACMG/AMP computational evidence across variant classes, model versions, and reference data
BibTeX
@misc{hingerl2026scooby2,
title = {{Scooby-2: fine-tuning AlphaGenome for single-cell genomic track prediction}},
author = {J. C. Hingerl and J. Gagneur},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.28.762238},
url = {https://www.biorxiv.org/content/10.64898/2026.10.28.762238v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Scooby-2: fine-tuning AlphaGenome for single-cell genomic track prediction
AU - J. C. Hingerl
AU - J. Gagneur
PY - 2026
DA - 2026/10/28
PB - bioRxiv
DO - 10.64898/2026.10.28.762238
UR - https://www.biorxiv.org/content/10.64898/2026.10.28.762238v1
AB - Sequence-to-function models predict genomic tracks from DNA sequence. AlphaGenome has been shown to outperform previous models, including Borzoi, on most track-prediction and variant-effect benchmarks. However, like its predecessors, it predicts tracks only for predefined, mostly bulk-derived cell types and tissues. While Scooby has extended Borzoi to single-cell resolution, a single-cell adaptation of AlphaGenome is still lacking. We present Scooby-2, which combines AlphaGenome with a cell-specific decoder to predict single-cell RNA-seq coverage and chromatin-accessibility genomic tracks from DNA sequence. The package provides the implementation, the parameter-efficient fine-tuning setup, the training pipeline and inference scripts. When evaluated on the NeurIPS bone marrow single-cell dataset, Scooby-2 achieved a Pearson correlation of 0.56 when predicting cell-type-specific gene expression differences, compared with 0.53 for Scooby. When using the distilled AlphaGenome backbone, variant effect prediction concordance of fine-mapped eQTLs from the OneK1K cohort improved from 0.37 to 0.47 Spearman's rank correlation across all associations, an improvement consistent with that of AlphaGenome over Borzoi itself.
ER -