Mutation processes leave different signatures in genes. Previous studies suggested that both the mutated and flanking bases influence somatic mutation characteristics. However, the understanding of how flanking sequences influence somatic mutation characteristics is limited. Here, we constructed A long short-term memory – self organizing map (LSTM-SOM) unsupervised neural network. By extracting mutated sequence features via LSTM and clustering similar features with SOM, we obtained 10 classes of mutant sequences (named mutation blots, MBs) from 2,141,527 somatic mutations in The Cancer Genome Atlas (TCGA) database. Differences features were revealed among MBs. MBs were related to clinical features, including age, sex, and cancer stage. Different kinds of MBs for specific genes may affect patient survival. Finally, we clustered the patients into 7 classes by MB composition. Significant differences in survival and clinical features were observed among different patient classes. This study provides a novel method for understanding mutant sequences and revealing the extensive relationships among mutant sequences, clinical features, and cancer patient survival.