1.7

1.7 (September 1994) is a minor revision from 1.6, incorporating a few new features. The following list gives a brief account of the new features.

• Custom priors. hmmb and hmmt now take an argument -p <priorfile>. If you don't like the Haussler prior Dirichlet distributions you can now provide your own.
• Dayhoff mixture priors. (HMM/profile hybrids) hmmb now takes an argument -P <PAM matrix file>. Symbol emission priors are then calculated from PAM scores instead of using a simple Dirichlet. This gives the ability to make an HMM equivalent of Gribskov/Eisenberg/McLachlan ``profiles''. For small numbers of sequences in the alignment, this makes more sensitive and discriminative models than the original prior.
• Ragged alignments. A new -R option to hmmb tells hmmb that the alignment is deliberately ragged-ended; hmmb then ignores external gaps, and does not count them as delete states. A new -R option to hmma telles hmma that the sequences to be aligned may include sequence fragments; hmma then uses Smith/Waterman instead of Needleman/Wunsch style alignment to produce the multiple alignment. hmma -R may remove the ends of some sequences, if they don't seem to match the model.
• Annotated models. In order to allow better viewing of HMM alignments, hmmb now picks up some annotation from SELEX-format alignments and keeps it with the model, permitting some annotation of consensus structure and/or landmark columns on the alignment. -F ``fancy output'' alignments produced by hmms,hmmsw, etc. align and print this annotation in addition to the previous alignment information. (does hmma?)
• Hypertext documentation. The User's Guide was expanded, and is now available in hypertext form on the World Wide Web.
• some minor reported bugs were fixed