hmm2pfam(1)

💖 hmm2pfam(1) — the HMMER Manual, but make it make sense 💖

okay bestie, gather round, because we're talking about hmm2pfam, one of the HMMER girlies, and she's lowkey iconic. this is the manual page for her, straight from the HMMER Manual. no cap, we're going through every single flag, every single default, every single vibe. let's go 🦋

---

✨ NAME ✨

hmm2pfam — search one or more sequences against an HMM database 💅

---

🎀 SYNOPSIS 🎀

hmm2pfam [options] hmmfile seqfile

---

🌸 DESCRIPTION 🌸

so here's the tea, bestie: hmm2pfam reads a sequence file called seqfile and compares each sequence in it, one at a time, against all the HMMs in hmmfile, looking for significantly similar sequence matches. it's giving "speed dating but for protein models" fr fr 👑

now, where does she even find hmmfile? she looks in the current working directory first, then in a directory named by the environment variable HMMERDB. the whole point is that admins can install HMM library(s) like Pfam in a common location and everyone can just vibe. 🫶

there's a separate output report for each sequence in seqfile. and that report is split into three sections:

1. a ranked list of the best scoring HMMs
2. a list of the best scoring domains, in order of their occurrence in the sequence
3. alignments for all the best scoring domains

now listen, because this part is a little sneaky: a sequence score may be higher than a domain score for the same sequence if there's more than one domain in that sequence. the sequence score takes into account all the domains. it's the group project of scoring, honestly.

every sequence scoring above the -E and -T cutoffs shows up in that first list, and then every domain found in this list shows up in the second list of domain hits. and if you want extra control? you can apply E-value and bit score thresholds to the <em>domain</em> list too, using the --domE and --domT options. slay. ✨

---

💅 OPTIONS 💅

• -h — print brief help. includes the version number and a summary of all the options, expert options included. your little cheat sheet, bestie. 📝
• -n — specify that models and sequence are nucleic acid, not protein. other HMMER programs autodetect this, but because of the order in which hmm2pfam accesses data, it can't reliably determine the correct "alphabet" by itself. so you gotta tell her. 🙈
• -A <n> — limits the alignment output to the <n> best scoring domains. and -A0 shuts off the alignment output entirely, which you can use to reduce the size of output files. -A0 is the "i'm just here for the numbers" era. 💀
• -E <x> — set the E-value cutoff for the per-sequence ranked hit list to <x>, where <x> is a positive real number. the default is 10.0. hits with E-values better than (less than) this threshold will be shown. lower = stricter, periodt.
• -T <x> — set the bit score cutoff for the per-sequence ranked hit list to <x>, where <x> is a real number. the default is negative infinity; by default the threshold is controlled by E-value and not by bit score. hits with bit scores better than (greater than) this threshold will be shown. 🔥
• -Z <n> — calculate the E-value scores as if we had seen a sequence database of <n> sequences. the default is arbitrarily set to 59021, which is the size of Swissprot 34. yes, an arbitrary number, and yes, she said it out loud. iconic.

---

👑 EXPERT OPTIONS 👑

okay these are the big girl options. same rules though: every detail stays. 💜

• --acc — report HMM accessions instead of names in the output reports. useful for high-throughput annotation, where the data are being parsed for storage in a relational database. database girlies, this one's for you. 🍓
• --compat — use the output format of HMMER 2.1.1, the 1998-2001 public release; provided so 2.1.1 parsers don't have to be rewritten. nostalgia core fr fr. 🌈
• --cpu <n> — sets the maximum number of CPUs that the program will run on. the default is to use all CPUs in the machine. overrides the HMMER_NCPU environment variable. only affects threaded versions of HMMER (the default on most systems). she will eat up your whole machine by default, and honestly? respect. 🔥
• --cut_ga — use Pfam GA (gathering threshold) score cutoffs. equivalent to --globT <GA1> --domT <GA2>, but the GA1 and GA2 cutoffs are read from each HMM in hmmfile individually. hmm2build puts these cutoffs there if the alignment file was annotated in a Pfam-friendly alignment format (extended SELEX or Stockholm format) and the optional GA annotation line was present. if these cutoffs are not set in the HMM file, --cut_ga doesn't work. so like... no GA line, no GA cutoff. that's the deal. 💅
• --cut_tc — use Pfam TC (trusted cutoff) score cutoffs. equivalent to --globT <TC1> --domT <TC2>, but the TC1 and TC2 cutoffs are read from each HMM in hmmfile individually. hmm2build puts these cutoffs there if the alignment file was annotated in a Pfam-friendly alignment format (extended SELEX or Stockholm format) and the optional TC annotation line was present. if these cutoffs are not set in the HMM file, --cut_tc doesn't work. same energy, different letters. ✨
• --cut_nc — use Pfam NC (noise cutoff) score cutoffs. equivalent to --globT <NC1> --domT <NC2>, but the NC1 and NC2 cutoffs are read from each HMM in hmmfile individually. hmm2build puts these cutoffs there if the alignment file was annotated in a Pfam-friendly alignment format (extended SELEX or Stockholm format) and the optional NC annotation line was present. if these cutoffs are not set in the HMM file, --cut_nc doesn't work. the trilogy is complete. 🎀
• --domE <x> — set the E-value cutoff for the per-domain ranked hit list to <x>, where <x> is a positive real number. the default is infinity; by default, all domains in the sequences that passed the first threshold will be reported in the second list, so that the number of domains reported in the per-sequence list is consistent with the number that appear in the per-domain list. consistency queen. 👑
• --domT <x> — set the bit score cutoff for the per-domain ranked hit list to <x>, where <x> is a real number. the default is negative infinity; by default, all domains in the sequences that passed the first threshold will be reported in the second list, so that the number of domains reported in the per-sequence list is consistent with the number that appear in the per-domain list.

important note, and i mean important: only one domain in a sequence is absolutely controlled by this parameter, or by --domT. the second and subsequent domains in a sequence have a de facto bit score threshold of 0 because of the details of how HMMER works. HMMER requires at least one pass through the main model per sequence; to do more than one pass (more than one domain), the multidomain alignment must have a better score than the single domain alignment, and hence the extra domains must contribute positive score. see the Users' Guide for more detail. 😭

• --forward — use the Forward algorithm instead of the Viterbi algorithm to determine the per-sequence scores. per-domain scores are still determined by the Viterbi algorithm. some have argued that Forward is a more sensitive algorithm for detecting remote sequence homologues; my experiments with HMMER have not confirmed this, however. so like... the discourse exists. the receipts do not. 🙈
• --informat <s> — assert that the input seqfile is in format <s>; do not run Babelfish format autodection. this increases the reliability of the program somewhat, because the Babelfish can make mistakes; particularly recommended for unattended, high-throughput runs of HMMER. valid format strings include FASTA, GENBANK, EMBL, GCG, PIR, STOCKHOLM, SELEX, MSF, CLUSTAL, and PHYLIP. see the User's Guide for a complete list. tell her the format, bestie, she'll stop guessing. 💖
• --null2 — turn off the post hoc second null model. by default, each alignment is rescored by a postprocessing step that takes into account possible biased composition in either the HMM or the target sequence. this is almost essential in database searches, especially with local alignment models. there is a very small chance that this postprocessing might remove real matches, and in these cases --null2 may improve sensitivity at the expense of reducing specificity by letting biased composition hits through. so it's a trade, and you should know you're trading. ⚖️
• --pvm — run on a Parallel Virtual Machine (PVM). the PVM must already be running. the client program hmm2pfam-pvm must be installed on all the PVM nodes. the HMM database hmmfile and an associated GSI index file hmmfile.gsi must also be installed on all the PVM nodes. (the GSI index is produced by the program hmm2index.) because the PVM implementation is I/O bound, it is highly recommended that each node have a local copy of hmmfile rather than NFS mounting a shared copy. optional PVM support must have been compiled into HMMER for --pvm to function. it's giving "cluster era" and honestly, we love a legacy moment. 🦋
• --xnu — turn on XNU filtering of target protein sequences. has no effect on nucleic acid sequences. in trial experiments, --xnu appears to perform less well than the default post hoc null2 model. so like... it's there. that's the review. 💀

---

🌈 SEE ALSO 🌈

Master man page, with full list of and guide to the individual man pages: see hmmer2(1).

for complete documentation, see the user guide (ftp://selab.janelia.org/pub/software/hmmer/2.3.2/Userguide.pdf); or see the HMMER web page, http://hmmer.janelia.org/. go read, bestie, it's free knowledge. 📚✨

---

💜 COPYRIGHT 💜

Copyright (C) 1992-2003 HHMI/Washington University School of Medicine. Freely distributed under the GNU General Public License (GPL). See the file COPYING in your distribution for details on redistribution conditions.

---

🫶 AUTHOR 🫶

Sean Eddy HHMI/Dept. of Genetics Washington Univ. School of Medicine 4566 Scott Ave. St Louis, MO 63110 USA http://www.genetics.wustl.edu/eddy/

---

okay that's the whole page, and you ate it, no crumbs. 💅 this was HMMER 2.3.2, dated Oct 2003, and you've been reading hmm2pfam(1). go forth and search those sequences, queen. 🎀🦋✨