Local sequence alignment

What is local sequence alignment?

Local sequence alignment is the process of finding and arranging the highest-scoring matching region within two sequences instead of comparing them end to end. Smith–Waterman dynamic programming allows an alignment to start and end internally under a substitution and gap model, making it appropriate for shared domains, motifs, fragments, or partial homology. A local score does not by itself provide aligned strings, coordinates, significance, or biological function.

Choose local alignment when only part of each sequence is expected to correspond. Examples include a conserved domain embedded in different architectures, a short motif, a sequencing fragment, or a query against longer candidates. If complete lengths should correspond, global alignment provides a more appropriate penalty model.

ProteinIQ’s StringZilla workflow returns Smith–Waterman score matrices, not traceback paths. Use those scores for rapid comparison and ranking, then use a method that emits alignments and coordinates when residue-level interpretation is required. Preserve scoring parameters because local scores are not comparable across arbitrary settings and lengths.

When to use local sequence alignment

• Best fit. Shared domains, motifs, fragments, and partial homology
• Required input. Sequences expected to share only a subsection and explicit scoring parameters

Benefits of local sequence alignment

• Clear correspondence. Ignores unrelated flanking regions
• Connected evidence. Finds partial similarity
• Reusable output. Supports fast pair prioritization

Primary limitations

• Method dependence. Score-only output lacks traceback
• Input dependence. Short motifs can occur by chance
• Interpretive limit. Significance needs additional context

Local sequence alignment methods

Smith–Waterman resets negative dynamic-programming cells to zero, so the best path can begin and end inside the matrix. The maximum cell gives the optimal local score under the selected parameters.

Database-search methods add heuristics, composition handling, and statistical calibration for large searches. A raw all-pairs score matrix is useful, but it should not be described as an E-value or a complete homology search.

Local sequence alignment applications

Local sequence alignment is best suited to shared domains, motifs, fragments, and partial homology. The result can support comparative review, sequence curation, annotation, profile construction, phylogenetic preparation, structural interpretation, or experimental planning when those downstream uses match the alignment scope.

Keep the alignment as evidence rather than a conclusion. Downstream claims should remain tied to sequence provenance, coverage, method agreement, relevant biological context, and any independent structural, evolutionary, or experimental support.

How to run local sequence alignment online

Use the connected workflow to keep input records, method settings, native outputs, warnings, and exports together. Review every stage before using the result for annotation, phylogeny, variant interpretation, or experimental decisions.

1. Define local question. Identify why only a subsection, rather than each full length, should correspond.
2. Prepare sequences. Remove avoidable low-complexity or boundary artifacts and verify molecule type.
3. Set score model. Record match, mismatch, gap, and normalization choices.
4. Calculate scores. Calculate the Smith–Waterman local score matrix for prioritization.
5. Trace selected hits. Use an alignment-producing method for selected pairs that need coordinates.

How to interpret local sequence alignment results

Compare scores only under the same algorithm and settings. Sequence length and composition affect raw values, and repetitive or low-complexity segments can create high local scores.

For selected pairs, inspect aligned coordinates, coverage, substitutions, gaps, and biological context. A strong local match may represent one shared domain rather than full-protein homology.

How local sequence alignment works

StringZilla calculates all-pairs Smith–Waterman local scores and explicitly returns scores rather than alignment paths.

1. Define local question. Identify why only a subsection, rather than each full length, should correspond.
2. Prepare sequences. Remove avoidable low-complexity or boundary artifacts and verify molecule type.
3. Set score model. Record match, mismatch, gap, and normalization choices.
4. Calculate scores. Calculate the Smith–Waterman local score matrix for prioritization.
5. Trace selected hits. Use an alignment-producing method for selected pairs that need coordinates.

Inputs and outputs

Check formats before running, then inspect and download the result from every workflow step.

Inputs

• Alignment input. FASTA PDB mmCIF Protein or nucleotide FASTA sequences for all-pairs local scoring.

Outputs

• Alignment outputs. FASTA CSV TSV PDB JSON Smith–Waterman local score matrices and downloadable tabular results; no traceback paths.