Information derived from multiple alignments of protein sequences can improve detection of distant relationships. We have found that blocks, which are ungapped multiple alignments representing the most highly conserved regions of proteins, are useful for efficiently utilizing this information. Blocks-based tools include the Blocks Database for protein classification, BLOSUM substitution matrices for scoring pairwise alignments, position-based sequence weights and pseudocounts for scoring multiple alignments, Blockmaker for motif identification, blocks-versus- blocks (LAMA) searching for detecting subtle similarities between families, and block-embedding methods for detecting distant relatives in sequence databanks. This talk will focus on embedding methods, which extract multiple alignment information from motif regions while retaining single sequence information where alignment is uncertain.