A new version of the database is live, covering papers through September 2026. v1.1.1 is now the default for every search type — SNV, genomic variants, gene fusions, CNVs and SVs. v1.0.0 remains selectable for back-tracking.
| Variant type | v1.0.0 | v1.1.1 | Change |
|---|---|---|---|
| SNV | 579,844 | 866,141 | +49.4% |
| Genomic variants (annotated rows) | 909,224 | 1,399,443 | +53.9% |
| Genomic variants (unique alleles) | 816,002 | 1,217,126 | +49.2% |
| Gene fusion | 85,966 | 142,948 | +66.3% |
| CNV | 29,457 | 83,101 | +182.1% |
| SV | 69,993 | 104,310 | +49.0% |
| Variant type | Unchanged | Renumbered | Split | Retired |
|---|---|---|---|---|
| SNV | 574,027 | 5,011 | 3 | 803 |
| Gene fusion | 85,966 | 0 | 0 | 0 |
| CNV | 29,360 | 0 | 0 | 97 |
| SV | 69,462 | 0 | 0 | 531 |
To enhance the user experience while searching and using the website, the display and search functions have been updated. Some tabs have been merged to make it simpler, and an Advanced search option is now available to search multiple terms at the same time (up to 5 terms). Please try it out!
A bug affecting text-field searches in the Genomic dataset — disease, phenotype and literature terms returned no results — has also been fixed.
The database itself is unchanged — but stay tuned! A new version of PubMind-DB, covering papers through September 2026, is coming shortly.
Please cite us at: Wang, P. & Wang, K. (2026). PubMind: literature-based genetic variant extraction and functional annotation using large language models. Nature Communications. Published online August 20, 2026. https://doi.org/10.1038/s41467-026-76834-4
To support reproducibility, transparency, and version tracking, the first public release of PubMind-DB is designated PubMind-DB v1.0.0. This database release was generated using PubMind v0.1.0, the first public release of the PubMind framework.
Reference: Wang, P. & Wang, K. (2026). PubMind: literature-based genetic variant extraction and functional annotation using large language models. Nature Communications. Published online August 20, 2026. https://doi.org/10.1038/s41467-026-76834-4
Provide 2 additional columns 'paper_level_pathogenicity_score' and 'pathogenicity_score_per_paper', which provide parathogenicity score averaging across paper in paper-level. Instead of the previous 'pathogenicity_score' which is paragraph-level averaging across all records.
Updated: April 22, 2026Update the Coordinate Search result page to provide a hyperlink directly to ClinVar, based on ClinVar AlleleID (CLNALLELEID).
Be careful about the Variation ID and Allele ID in ClinVar, as they are different. The ID in PubMind is Allele ID, but the ID in ClinVar URL and webpage is Variation ID. The ClinVar version used in PubMind-DB is 20240917.
The website has been updated to support more search functions and improved user interface.
(1) All fields support multiple-term search now — separate each term by a new line.
(2) The Identifier Search allows users to search a variant based on the cDNA or protein change, e.g. BRCA1:c.A122T, BRCA1:p.Arg133Cys
(3) The result page lets users customize which columns to display.
(4) Added genomic variant search under Coordinate Search (VCF-style, e.g. chr17_43100863_C_T).
The latest version is presented at the American Society of Human Genetics (ASHG) 2025 Annual Meeting. The website interface and search functions are still in development, but the content (database itself) is fixed and notification will be sent if database is changed.
Updated: October 16, 2025