Figure 3.

Hybrid NP models using CPMs and highly predictive HPTMs have higher imputation accuracy than using only CPMs

(A and B) Average association strength for each HPTM in the (A) acetylation and (B) methylation panels.

(C) Summary of hybrid NP models that use CPMs and cleaved H3T22 (acetylation panel) or H3K36me1 (methylation panel) as inputs to impute an HPTM and comparison with models in task 1 (CPMs → HPTM). Each dot represents a unique model. Lines connect models predicting the same HPTM. One-sided, paired Wilcoxon test was used to compute the significance of the improvement in R² due to the hybrid models (n = 36 HPTMs per task).

See also Figure S7.