PfamB

PfamB is a component of the Pfam database that complements the curated Pfam-A families. It consists of automatically generated protein family clusters derived from sequences that do not align to any Pfam-A family. The primary purpose of PfamB is to provide additional seed alignments and models that may hint at distant or novel relationships among proteins, offering a starting point for discovering new families.

PfamB entries are produced through automated clustering of non-Pfam-A UniProt sequences, followed by the construction of profile hidden Markov models (HMMs) from the resulting alignments. Because these families are created automatically, they are not subjected to the same level of manual curation as Pfam-A. As a result, PfamB models are generally considered less robust and reliable for precise annotation, and users are advised to interpret hits with caution and to seek corroborating evidence when possible.

In practice, PfamB serves as a reservoir of potential new families that could be promoted to Pfam-A upon further validation and curation. The balance between coverage and accuracy is a key consideration in using PfamB, and researchers often use it as a supplementary resource alongside Pfam-A rather than as a primary basis for functional assignment.

PfamB has played a historical role in expanding the scope of Pfam by capturing sequence diversity not yet represented in Pfam-A. Over time, developments in sequence data and analysis methods have influenced how PfamB is updated and integrated, with an emphasis on critical evaluation before inclusion as a curated Pfam-A family.