How to Use HMMER Profile Alignment

HMMER Profile Alignment uses the hmmalign program from HMMER to align biological sequences against a profile hidden Markov model. Unlike general-purpose multiple sequence alignment methods that infer relationships directly among all submitted sequences, hmmalign aligns each sequence independently to the same profile. The shared profile coordinates then induce a multiple sequence alignment anchored to the model's consensus match states.

This workflow is well suited to adding newly discovered homologs to an established protein family, aligning domain fragments to a curated Pfam-style model, preparing profile-consistent inputs for phylogenetic analysis, or maintaining stable consensus coordinates as a sequence collection grows. It supports protein, DNA, and RNA profiles, but each job must use one alphabet consistently across the uploaded profile and all input sequences.

hmmalign uses unihit local alignment. Input sequences should already be known to contain one copy, or a fragment, of the modeled domain. It is not a database-search or homology-detection service, and it does not identify repeated domains in a long sequence. Use HMMER search tools first when the locations or presence of profile matches are unknown.

A profile HMM represents a sequence family's position-specific residue preferences together with insertion and deletion probabilities. hmmalign finds an optimal path for each input sequence through that model. Residues assigned to profile match states become homologous consensus columns, while residues assigned to insert states remain insertions relative to the profile rather than confidently aligned positions.

That distinction appears directly in the output notation. Uppercase residues occupy profile match columns, lowercase residues are insertions relative to the profile, hyphens mark deletions in match columns, and periods pad insertion columns. The Stockholm output also retains HMMER posterior-probability annotations and an RF reference line identifying profile consensus columns. Researchers should preserve those annotations when confidence-aware masking or profile-coordinate mapping matters.

Profile HMM must be a readable HMMER profile. If it contains several models, hmmalign uses the first one. Input Sequences may be proteins, DNA, or RNA, but mixed alphabets are rejected. The sequences should represent single-domain hits or fragments; unrelated sequence, alphabet mismatches, or inconsistent domain multiplicity can produce misleading alignments or a failed run.

Model Alignment is optional and corresponds to hmmalign's map alignment behavior. Supply it only when it is exactly the alignment used to build the uploaded profile. HMMER then reproduces its consensus columns and adds the new sequences in the same profile coordinate system. Trim Terminal Residues removes residues assigned to nonhomologous N- and C-terminal states; leaving it disabled retains those residues at the alignment edges.

The service exports alignment.sto for full Stockholm annotations, alignment.fasta for broad downstream compatibility, and alignment.csv for table-based analysis. Because insert-state residues are not mutually aligned in the same sense as match-state residues, apparent offsets among lowercase insertions should not automatically be interpreted as alignment errors.

• Stable profile coordinates: Add new sequences without rebuilding a de novo alignment or shifting the family's consensus coordinate system.
• Protein and nucleotide workflows: Use the same managed service for amino-acid, DNA, and RNA profile HMMs.
• Confidence-aware output: Retain Stockholm posterior probabilities and reference-column annotations for masking and downstream interpretation.
• Curated-family expansion: Include the original model-building alignment when new sequences must be added to an established seed alignment.
• Interoperable results: Download Stockholm, aligned FASTA, and CSV representations for phylogenetics, annotation, visualization, or custom analysis.