Skip to main content
Fig. 6 | Genome Medicine

Fig. 6

From: Mutation-Attention (MuAt): deep representation learning of somatic mutations for tumour typing and subtyping

Fig. 6

Association between MuAt-derived mutation motifs and tumour types. Attention values for mutation motif pairs (X-axis) extracted from the MuAt model trained on PCAWG data. Values have been averaged first over tumours and then over tumour types (Y-axis). Types of the key and query mutations (SNV, MNV, indel or SV) are indicated on the two top rows. Every fourth motif pair is labelled as “X_Y” where X and Y are the motifs corresponding to attention query and key, respectively. Label colours indicate mutation types. Motif groups: A SNV/SNV and B SNV/non-SNV pairs appearing in many tumour types. C Motifs with doublet C>T substitutions specific to skin melanomas. D and E Motifs characteristic to chronic lymphocytic leukemias and Non-Hodgkin lymphomas

Back to article page