Fig. 5From: Achieving robust somatic mutation detection with deep learning models derived from reference data sets of a cancer samplePerformance comparisons on the library preparation and FFPE data sets. a, b F1-score (%) comparison for different somatic mutation callers across different library kits and DNA amounts. Results of different callers for the same library preparation kit and DNA amount are connected for better illustration. c F1-score (%) comparison for different somatic mutation callers on 16 FFPE WGS replicates with FFPE or fresh matched normal. d SNV F1-score (%) comparison for different somatic mutation callers on 14 FFPE WES replicates with FFPE or fresh matched normal. Here, the SEQC-WGS-GT50-SpikeWGS10 trained models were used for NeuSomatic and NeuSomatic-SBack to article page