Description
Ranks hmmer/hmmsearch hits by score/e-value and, optionally, attaches per-hit domain coordinates -- reading the already-parsed, already-typed Parquet table(s) hmmer/formattsv
- duckdb/table2parquet produce, rather than parsing raw hmmsearch tblout/domtblout text itself. The ranking and domain-coordinate merge run as DuckDB SQL (window functions, disk-spillable temp tables) instead of in-memory R/dplyr, so memory use no longer scales with the full width of every input row held at once.
Tools
hmmer Documentation
Biosequence analysis using profile hidden Markov models
duckdb Documentation
An in-process SQL OLAP database management system
https://github.com/duckdb/duckdbLicense: MIT
Command copied!