Pooled combinatorial screening identifies transcription factor sets that drive hematopoietic progenitor-like cell fate

Mammalian cells can be directed toward specific fates by overexpression of transcription factors (TFs). However, discovering and optimizing which TFs in combination produce a state of interest remains challenging. Here, we develop a scalable screening platform that addresses this challenge by combining high multiplicity of infection (MOI), pooled delivery of barcoded TF open reading frames (ORFs), data augmentation, targeted cell enrichments, and single-cell transcriptomic readouts. As proof of principle, we apply the platform to optimize the generation of hematopoietic stem and progenitor-like cells (HSPCs) from human embryonic stem cells. Our data demonstrate technical performance across a range of key metrics and reveal a richly structured reprogramming fitness landscape over millions of TF combinations. In silico optimization of HSPC-similarity metrics over this landscape revealed two TF combinations that demonstrate superior potency in generating naive multipotent hematopoietic progenitors relative to gold-standard controls. This study demonstrates a powerful approach for data-driven cell fate engineering using complex combinatorial perturbations.