Fig. 5. Mouse temporal signatures are predictive of histology state in a cohort of inflammatory bowel disease patients.

Comparison of adjusted R squared values for penalized ordinary least squares models of histological measurements using human orthologs of DSS disease evolution signatures and combined Nancy (UC) and GHAS (CD) signatures in a) (non-Ileal) patient tissue samples and b) patient blood samples. Each model fit the first five principal components of each signature (including splines with three knots to test for non-linear associations), penalized by extremizing the Akaike Information Criterion (AIC), and bias-corrected R-squared values were computed via 300 rounds of bootstrap resampling with replacement. Tissue specific histological scores per patient were averaged across multi-regional tissue samples.