Synthetic Sequences Supporting "Preference-Based Fine-Tuning of Genomic Sequence Models for Personal Expression Prediction with Data Augmentation"

This dataset contains synthetic genomic sequences used in the study“Preference-Based Fine-Tuning of Genomic Sequence Models for Personal Expression Prediction with Data Augmentation”(bioRxiv preprint: https://doi.org/10.1101/2025.11.09.687505).

🔹 Synthetic Sequences

To create a diverse training set for the study, we generated 1,000 synthetic individuals using the sim1000G framework (Dimitromanolakis et al., BMC Bioinformatics 20(1), 26, 2019).sim1000G simulates recombination and mutation based on 1000 Genomes haplotypes, producing realistic but non-identifiable variant structures.

All sequences included in this dataset:

are synthetic and derived solely from simulated genotypes,

are uppercase-normalized (A/C/G/T/N),

can be used to reproduce the methodological components of the study (model training, data augmentation, preference-based fine-tuning).

Because identifiable human genomic sequences cannot be redistributed, only these synthetic sequences are publicly released.

🔹 Real Sequences

This dataset does not contain real haplotype sequences from the GEUVADIS cohort (E-GEUV-1).However, users may reconstruct the personalized 196,608 bp windows (±98,304 bp around TSS), following Huang et al. (2023), by extracting the corresponding hg19 reference intervals and substituting sample-specific variants.

A full, scripted reconstruction pipeline is available at:

https://github.com/pacifiic/augment-finetune-genomics

The original GEUVADIS WGS and RNA-seq data for 421 phased individuals remain accessible through the European Bioinformatics Institute (EBI):

https://www.ebi.ac.uk/biostudies/arrayexpress/studies/E-GEUV-1

For preprocessing, phasing, and window-construction details, please consult the GitHub repository and the reference paper.