By integrating bulk and single-cell multiomics at unprecedented scale, PertFormer achieves zero-shot prediction of causal perturbation effects and cell fate transitions—making in silico regulatory discovery feasible where wet-lab experiments were previously the only option.
PertFormer, a 3-billion parameter foundation model pretrained on multimodal bulk and single-cell genomic data, enables zero-shot prediction of functional perturbation effects, gene regulatory networks, and cell fate transitions without task-specific fine-tuning, outperforming existing methods by 20.9%-480.2% and identifying experimentally validated novel cancer therapeutic targets through interpretable attention-based simulation of genomic and epigenomic perturbations.
Deciphering cell type-specific perturbation effects on genes and cellular states demands considerable experimental resources. To overcome this challenge, we propose Perturbation Transformer (PertFormer), a foundation model to uncover functional and regulatory mechanisms through interpretable in silico perturbation process. PertFormer comprises 3 billion parameters pretrained on both bulk and single-cell datasets covering 9 types of multiomics (55 billion bp bulk multiomics and the largest-to-date 1.5 billion paired multiomic samples from 1 million single cells), capturing regulation of 300 kb around every genic region. PertFormer enables zero-shot prediction of functional perturbations and cell fate dynamics, outperforming existing methods by 20.9%-480.2%. PertFormer identifies novel tumor treatment targets, validated experimentally. The generalization capabilities of PertFormer have the potential to accelerate the discovery of biological and clinical targets.