Delineating transcriptomic signatures of in vitro human skeletal muscle models in comparison to in vivo references

A pivotal question at the heart of stem cell research is how faithful cellular models recapitulate human tissues. Skeletal muscle, the largest organ in the human body, has been modeled by various in vitro systems. Here, we sought to delineate the state-of-the-art of muscle models by performing a large-scale analysis of transcriptome datasets, covering over 400 samples across 39 studies, including bulk and single-cell RNA sequencing of 2D and 3D models and their in vivo counterparts. By comparing these models to in vivo muscle, we highlighted failed upregulation of myogenic factors and retention of epigenetic memory from the in vitro source material. We featured differences in lipid metabolism and depletion of multiple fibroblast growth factor (FGF) ligands in the in vitro models. Finally, we revealed model-dependent variation in myogenic progenitors. Our analyses highlight targetable processes to improve the models while paving the way for similar studies on other cell types.