Extended Data Table 1: Implications of this study for the use of cell lines in cancer research.

A summary of the main findings of this study, their practical implications, and recommendations for addressing them.

Findings	Implications	Recommendations
Given two strains, ~20% of mutations would be observed in only one of them	There is ~10% likelihood that a mutation observed in a strain would not appear in a database of cell line genomic features	• Be cautious when using published datasets of genomic features as “lookup tables”
Prolonged passaging introduces more variation than multiple freeze-thaw cycles	For most cell lines, freezing and thawing is likely to be associated with fewer changes than maintaining in culture	• Keep track of passage number • Use passage-matched controls • For large-scale screens, prepare multiple frozen vials for downstream analyses
Various genomic, transcriptomic and phenotypic assays yield highly similar clustering trees	Simple and inexpensive genome-wide assays can serve as a proxy for diversification	• Use inexpensive genome-wide assays (e.g., LP-WGS) and compare to published references using Cell STRAINER: https://cellstrainer.broadinstitute.org • Exclude strains that show extreme diversification
Genetic manipulations that are considered “neutral” can introduce genetic variation	Cell lines with fluorescent reporters, DNA barcodes or Cas9 expression are not identical to their parental cell lines	• Use efficient infection methods to reduce the bottleneck associated with antibiotic selection • Characterize manipulated strains to ensure they retain hallmark genomic features • In CRISPR screens, correct for copy number effects using the copy number landscape of the screened strain
Genetic and transcriptomic variation may affect drug response	Inconsistencies in drug response studies may be attributed to genetic and transcriptomic variability	• Genetic and transcriptomic distances should be considered when comparing drug response data • Compare drug response data to genomic data from the same strain
	Transcriptional differences between sensitive and resistant strains can elucidate compound mechanism of action	• Use characterized isogenic-like strains to uncover associations between molecular features and drug response
Pre-existing heterogeneity within culture underlies cell line instability	Single cell-derived clones differ from one another genetically, transcriptionally and phenotypically	• Confirm the genomic features of single cell-derived clones • Avoid comparisons between bottlenecked cell populations, whenever possible
	Subtle differences in culture conditions can lead to changes in cell line clonal composition	• Keep culture conditions constant
Heterogeneity keeps emerging in culture due to ongoing genomic instability	Prolonged passaging of single cell-derived clones can lead to their diversification Cell lines with deficient maintenance of genome integrity (e.g., MSI or TP53-mutant) are more prone to genomic evolution	• Re-confirm genomic features of single cell-derived clones following prolonged passaging • Apply these recommendations more stringently to genomically unstable cell lines