SeekPCMdb: knowledge on disease-associated protein-coding mutations

Protein-coding mutations (PCMs) play crucial roles in the onset and progression of human diseases. With the widespread application of next-generation sequencing, an increasing number of disease-associated PCMs have been reported. However, these mutations are scattered across a vast body of literature, limiting their accessibility for systematic investigation. To address this challenge, we present SeekPCMdb (https://ngdc.cncb.ac.cn/seekpcmdb/), a literature-derived knowledge base that systematically catalogs human disease-related PCMs. SeekPCMdb employs a semi-automated workflow that integrates manual curation of high-quality publications with large language model (LLM)-assisted information extraction, capturing gene–mutation–disease relationships from >12 000 PubMed abstracts. The current release contains 11 395 curated entries, covering 3516 genes, 4505 diseases, and 23 mutation types. Each entry is richly annotated at multiple levels, including gene function, mutation consequence, and disease classification, enabling thorough biological interpretation. Comparative analysis shows that 63.7% of PCMs in SeekPCMdb are not cataloged in ClinVar, underscoring its unique contribution. SeekPCMdb features a user-friendly web interface that supports interactive searching, visualization, and download, along with multi-dimensional filtering based on gene, mutation type, or disease. By bridging unstructured literature and structured mutation knowledge, SeekPCMdb provides an invaluable resource for investigating genotype-phenotype correlations, assessing mutation pathogenicity, and identifying cross-disease mutational patterns.