The paper introduces a method called HHsearch for detecting distant homologous relationships between proteins using profile hidden Markov models (HMMs). The method generalizes the alignment of protein sequences with a profile HMM to the pairwise alignment of profile HMMs. HHsearch is benchmarked against other homology detection tools such as BLAST, PSI-BLAST, HMMER, PROF-SIM, and COMPASS in an all-against-all comparison of a database of 3691 protein domains from SCOP 1.63 with pairwise sequence identities below 20%. The results show that HHsearch is significantly more sensitive than these tools, detecting between 2.7 and 4.2 times more homologs than PSI-BLAST or HMMER and between 1.44 and 1.9 times more than COMPASS or PROF-SIM for a rate of false positives of 10%. Additionally, HHsearch produces higher-quality alignments, with 1.2, 1.7, and 3.3 times more good alignments at the family, superfamily, and fold levels, respectively. The method also includes a statistical method for scoring predicted versus known secondary structure, which further improves sensitivity. HHsearch is available for download and as a web server, providing a powerful tool for remote homology detection, functional annotation, structure prediction, and protein evolution studies.The paper introduces a method called HHsearch for detecting distant homologous relationships between proteins using profile hidden Markov models (HMMs). The method generalizes the alignment of protein sequences with a profile HMM to the pairwise alignment of profile HMMs. HHsearch is benchmarked against other homology detection tools such as BLAST, PSI-BLAST, HMMER, PROF-SIM, and COMPASS in an all-against-all comparison of a database of 3691 protein domains from SCOP 1.63 with pairwise sequence identities below 20%. The results show that HHsearch is significantly more sensitive than these tools, detecting between 2.7 and 4.2 times more homologs than PSI-BLAST or HMMER and between 1.44 and 1.9 times more than COMPASS or PROF-SIM for a rate of false positives of 10%. Additionally, HHsearch produces higher-quality alignments, with 1.2, 1.7, and 3.3 times more good alignments at the family, superfamily, and fold levels, respectively. The method also includes a statistical method for scoring predicted versus known secondary structure, which further improves sensitivity. HHsearch is available for download and as a web server, providing a powerful tool for remote homology detection, functional annotation, structure prediction, and protein evolution studies.