TIGRFAMs is a manually curated database of protein families described by hidden Markov models (HMMs) and associated information. It supports both automated and manual annotation of genomes. The database includes models for full-length proteins and shorter regions at the levels of superfamilies, subfamilies, and equivalogs. Equivalogs are sets of homologous proteins conserved in function since their last common ancestor. The scope of each model is determined by adjusting cutoff scores and selecting members of the seed alignment to group proteins with specific functions or general properties. TIGRFAMs is complementary to Pfam, which typically covers distant homologs but ends at the boundaries of conserved structural domains. The database currently contains over 1600 protein families. TIGRFAMs is available for searching or downloading at www.tigr.org/TIGRFAMs. TIGRFAMs includes models of type equivalog, subfamily, and domain. More than half of the models are of type equivalog. Each equivalog model provides a strong prediction of the specific protein function named by the model to any protein that scores above its trusted cutoff. TIGRFAMs also includes subfamily models, which can represent shared functions among proteins with varying functions and help avoid misannotation due to overinterpreting homologs as equivalogs. The database is designed to be complementary to Pfam, using the same HMMER package. TIGRFAMs version 2.1 contains 1622 families, of which 837 are classified as equivalogs. Coverage in newly sequenced bacterial genomes is estimated at 20% for all TIGRFAMs models and 10% for equivalog models. TIGRFAMs focuses on microbial proteins, with special attention to transporters, DNA repair proteins, and other areas. The database is supported by the Department of Energy. Users are invited to provide feedback and contributions.TIGRFAMs is a manually curated database of protein families described by hidden Markov models (HMMs) and associated information. It supports both automated and manual annotation of genomes. The database includes models for full-length proteins and shorter regions at the levels of superfamilies, subfamilies, and equivalogs. Equivalogs are sets of homologous proteins conserved in function since their last common ancestor. The scope of each model is determined by adjusting cutoff scores and selecting members of the seed alignment to group proteins with specific functions or general properties. TIGRFAMs is complementary to Pfam, which typically covers distant homologs but ends at the boundaries of conserved structural domains. The database currently contains over 1600 protein families. TIGRFAMs is available for searching or downloading at www.tigr.org/TIGRFAMs. TIGRFAMs includes models of type equivalog, subfamily, and domain. More than half of the models are of type equivalog. Each equivalog model provides a strong prediction of the specific protein function named by the model to any protein that scores above its trusted cutoff. TIGRFAMs also includes subfamily models, which can represent shared functions among proteins with varying functions and help avoid misannotation due to overinterpreting homologs as equivalogs. The database is designed to be complementary to Pfam, using the same HMMER package. TIGRFAMs version 2.1 contains 1622 families, of which 837 are classified as equivalogs. Coverage in newly sequenced bacterial genomes is estimated at 20% for all TIGRFAMs models and 10% for equivalog models. TIGRFAMs focuses on microbial proteins, with special attention to transporters, DNA repair proteins, and other areas. The database is supported by the Department of Energy. Users are invited to provide feedback and contributions.