Skip to main content
Glama

hmm_build

Read-only

Build a profile HMM from aligned sequences on this machine, as HMMER 3.4's hmmbuild builds one: columns whose weighted residue fraction clears symfrac become match states, rows are Henikoff position-based weighted so near-duplicates do not outvote a divergent member, the counts are entropy-weighted down to a target information per position, and the model is calibrated for E-values by hmmbuild's own simulation.

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "hmm_build"
  }
}

Input Schema

NameRequiredDescriptionDefault
EfLNohmmbuild --EfL: their length, in residues. Only applies when calibrate is true. Advanced.
EfNNohmmbuild --EfN: random sequences the Forward fit is made over. Only applies when calibrate is true. Advanced.
EftNohmmbuild --Eft: the tail mass the Forward exponential is fitted to. Only applies when calibrate is true. Advanced.
EmLNohmmbuild --EmL: their length, in residues. Only applies when calibrate is true. Advanced.
EmNNohmmbuild --EmN: random sequences the MSV fit is made over. Only applies when calibrate is true. Advanced.
EvLNohmmbuild --EvL: their length, in residues. Only applies when calibrate is true. Advanced.
EvNNohmmbuild --EvN: random sequences the Viterbi fit is made over. Only applies when calibrate is true. Advanced.
ereNohmmbuild --ere: the target mean relative entropy in bits per position. 0 takes hmmbuild’s alphabet default (0.59 protein, 0.62 nucleic). Only applies when effn is entropy.
effnNo`entropy` scales the counts until the model carries `ere` bits per position, which makes it a family model rather than a memory of this alignment; `none` uses the rows as counted; `set` uses `eset`. Values: entropy (Entropy-weighted), none (Actual count), set (Fixed (eset)).entropy
esetNohmmbuild --eset: the effective sequence count to use. Only applies when effn is set.
nameNothe model’s name, written on its NAME line.profile
seedNohmmbuild --seed: the simulation is reproducible, and the same seed gives the same statistics. Only applies when calibrate is true.
_filesNoThe project: record references shaped like _subject. Their records become library documents a tool names in sequences/parts/document; a trace or alignment is found by file name.
esigmaNohmmbuild --esigma: the entropy target’s small-model correction. 0 takes hmmbuild’s default. Only applies when effn is entropy. Advanced.
symfracNohmmbuild --symfrac: a column becomes a match state when this weighted fraction of rows carry a residue there rather than a gap.
_subjectNoThe record to act on: exactly one of sequence, accession (+source), sample (an id from list_samples) or file ({name, text} or {name, base64}); record_index picks one record of several.
alphabetNo`auto` reads the rows the way Easel does. Set it when a short alignment is read wrongly. Values: auto (Auto (Easel’s rule)), amino (Protein), DNA (DNA), RNA (RNA). Advanced.auto
alignmentNothe aligned rows as `{ name, aligned }`, gaps included: wire align_sequences’ `alignedSequences` here. Rows of different lengths are refused as unaligned.
calibrateNofit E-value statistics by simulation, as hmmbuild does. Off, the model has bit scores and no E-values, and hmmscan refuses it.
weightingNo`pb` is Henikoff position-based weighting (hmmbuild’s default), so twenty near-identical rows do not outvote one divergent member; `none` counts every row once. Values: pb (Position-based (Henikoff)), none (Equal).pb
fragthreshNohmmbuild --fragthresh: a row spanning less than this fraction of the alignment is a fragment, and its leading and trailing gaps are read as missing data rather than deletions.
alignment_fileNoor an alignment file, by name, from the project or the `_files` sent with the call: Stockholm, Clustal, PHYLIP, MSF or aligned FASTA, read by the same reader the alignment tab opens it with.

Output Schema

NameRequiredDescriptionDefault
hmmYesthe model as HMMER3/f text: wire it into hmm_search’s `profile_text`, or keep it as a `.hmm`
statsYesthe fitted MSV, Viterbi and Forward mu and lambda; absent when calibration was off
profileYesthe model built: name, match states, alphabet, effective sequence count, whether it carries E-value statistics
weightsYeseach row’s relative weight, in input order
fragmentsYesrows the fragment rule treated as partial sequences
matchColumnsYesthe alignment columns that became match states, 1-based
meanRelativeEntropyYesmean relative entropy per match state, in bits (hmmbuild’s re/pos)

Schema Changelog

Changes observed during successful MCP inspections.

  1. Added

TDQS

A3.9/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare readOnlyHint=true, destructiveHint=false and openWorldHint=false, and the description is consistent with a local, non-destructive computation ('on this machine'). Beyond that it discloses real mechanics — match-state assignment via symfrac, Henikoff weighting to prevent near-duplicate domination, entropy weighting to a target bits/position, and simulation-based E-value calibration with a reproducible seed — which tells an agent how outputs will behave. It stops short of stating cost/runtime implications of the simulation calibration, which is the main remaining gap.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

A single front-loaded sentence opens with the verb and resource, then uses a colon to enumerate the four algorithmic stages, so the reader can stop after clause one. Dense but every clause carries distinct information; no filler or restatement of the title.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 22-parameter tool with nested _subject/_files objects, the description covers the conceptual pipeline and reproducibility, and an output schema exists so return values need not be explained. What it omits is input-selection guidance among alignment / alignment_file / _subject and any indication of runtime weight from calibration, leaving a modest gap against the tool's complexity.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so the per-parameter baseline is 3, but the description adds conceptual glue the schema cannot: it explains how symfrac, weighting, effn/ere and calibrate interact to produce a family model rather than a memorized alignment. That raises it above baseline, though individual parameter meaning is largely already carried by the schema strings.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource — 'Build a profile HMM from aligned sequences on this machine' — and pins it to HMMER 3.4's hmmbuild semantics, so an agent knows exactly what operation is performed. It does not explicitly name a sibling (e.g. hmm_search) to contrast with, but build-vs-search is unambiguous from the verb and from the note that an uncalibrated model is refused by hmmscan.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Usage is only implied: the tool takes aligned rows, and the schema notes elsewhere say to wire align_sequences' `alignedSequences` here and that calibration off makes the model unusable by hmmscan. The description itself contains no explicit when-to-use/when-not guidance or stated prerequisites (e.g. that an alignment or alignment_file/_subject must actually be supplied, though no parameter is marked required).

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources