Induced pluripotent stem cells (iPSCs) are transforming cell biology, yet the inherent variability among iPSC lines and challenges in accurately characterizing iPSC-derived cell types have limited their broader use in preclinical screening. In this seminar, an innovative imaging assay is presented that combines cell painting with convolutional neural networks to reliably identify cell types even in dense, mixed cultures. The method is benchmarked using pure and mixed cultures of neuroblastoma and astrocytoma cell lines, achieving classification accuracies exceeding 96%. Furthermore, the approach is refined by focusing on the nuclear region and its immediate environment, which maintains high accuracy even under challenging, high-density conditions.

The targeted profiling strategy is further applied to iPSC-derived neural cultures to assess differentiation status by quantifying the ratio of postmitotic neurons to neural progenitors, with cell-based predictions outperforming traditional time-in-culture metrics (96% vs. 86%). In mixed iPSC-derived neuronal cultures, microglia are unequivocally discriminated from neurons, and a tiered analysis allows for further distinction between activated and non-activated microglial states, albeit with lower accuracy. Thus, morphological single-cell profiling is demonstrated as an effective means to monitor cell composition in complex neural cultures, holding great promise for enhancing quality control in iPSC-derived cell models for preclinical applications.