In order to determine if the observed aberrations in genomic imprints resulted from in vitro manipulations, we selected 140 hPSC samples for which we had detailed histories of culture media, passaging techniques and growth substrates (Table S7). For each imprinted gene (dependent variables), we generated two separate multiple linear regression models, which in addition to the source lab, considered each in vitro manipulation (independent variables) either in combination with other concurrent in vitro manipulations (model 1; e.g. number of manual passages in Wicell medium on MEFs) or in isolation (model 2; e.g. number of passages in Wicell medium). Analysis of these models showed apparent lab-specific effects for DIRAS3, L3MBTL and PEG3 (Bonferroni-adjusted p<0.001, Table S7A–C). We constructed correlation matrices to investigate inter-variable relationships that could potentially explain these apparent lab-specific effects (Table S7D–E). DIRAS3 aberrations were most highly correlated with the Laslett lab samples (R=0.59), which was the only lab positively correlated with use of the original medium used to maintain hESCs (“originalES”, R=0.63) (Thomson et al., 1998). This lab-specific effect can therefore be explained by the use of originalES
