6 Protein Bioinformatics

Protein bioinformatics and evidence

Protein bioinformatics uses sequence and structure data, databases, and computational methods to investigate protein functions, relationships, and molecular features. Computational results are hypotheses, so they should be interpreted alongside experimental evidence and the quality of the underlying data.

Sequences and protein features

A Protein sequence is a linear chain of amino acids, usually represented by one-letter symbols. Residue order constrains folding and helps determine properties such as charge, hydrophobicity, and potential chemical interactions.

Sequence analysis can identify or predict signal peptides and transit peptides that may direct a protein to a cellular destination; transmembrane segments, often rich in hydrophobic residues, that may anchor a protein in a membrane; and motifs and active or binding sites associated with biochemical roles. It can also identify or predict post-translational modifications, including phosphorylation or glycosylation, sites of proteolytic cleavage, and disordered regions that may be flexible and lack a single stable structure in isolation.

UniProtKB provides protein sequences with annotations and evidence. Some features are experimentally supported or manually curated, while others are computationally predicted. Check the evidence and annotation status before treating a feature as established.

Families, domains, and motifs

A Protein family groups evolutionarily related proteins. A Domain is a recognizable part of a protein sequence that can often fold or function as a unit, and a single protein may contain several domains. A Motif is a shorter conserved sequence pattern that may indicate a binding or catalytic site. These concepts are related but not interchangeable: a short motif alone usually provides less evidence for a specific function than a well-supported family or domain match.

InterPro integrates predictive signatures from member databases to classify proteins into families and identify domains, repeats, and important sites. InterProScan applies these models to submitted sequences. The domain architecture—the order and combination of domains—can help distinguish proteins with different functions even when they share one domain.

Alignment and similarity searches

An alignment places residues from two or more sequences into corresponding positions while allowing substitutions and gaps. A Global alignment compares sequences across their full lengths and is useful for sequences of similar length and coverage. A Local alignment finds the most similar matching region and is useful when sequences differ in length or share only one conserved region.

Alignment algorithms assign scores using substitution matrices, such as BLOSUM, and gap penalties. A database search such as BLASTP can find related proteins and report an E-value: the expected number of matches of similar score that could arise by chance in a database of that size. Smaller E-values generally provide stronger evidence of sequence similarity, but interpretation also depends on sequence length, database size, coverage, and biological context. Similarity supports, but does not by itself prove, common ancestry or identical function.

For a small illustrative local alignment in Python, Biopython’s PairwiseAligner supports substitution matrices and configurable gap penalties:

The score ranks alignments under the chosen settings; it is not a direct probability that the proteins share a function. Real analyses require biologically suitable sequences and scoring parameters, and database searches when appropriate.

Structures and confidence

The Protein Data Bank (PDB) is an archive of experimentally determined three-dimensional structures of biological macromolecules, with coordinates and associated experimental and descriptive information. A structure can show how a protein binds a ligand, forms a complex, or positions residues in an active site. However, it may represent only part of a protein or one particular experimental condition.

The AlphaFold Protein Structure Database provides computationally predicted structures. Treat them as models, not experimental observations. Inspect confidence information: pLDDT estimates local confidence residue by residue, while predicted aligned error (PAE) helps assess uncertainty in the relative placement of regions. A high-confidence local fold does not automatically establish the arrangement of separate domains, a protein interaction, or a biological mechanism.

A workflow for protein analysis

1. Identify the sequence. Obtain the amino-acid sequence and a stable identifier from a resource such as UniProtKB. Confirm the organism and whether the entry is reviewed; check for missing residues, ambiguous symbols, or isoforms.

2. Search for relatives. Use a protein similarity search to find homologous sequences. Examine alignment coverage, identity, conserved regions, and statistical significance—not just the top hit.

3. Annotate features and domains. Review curated sequence annotations and scan the sequence with InterPro or InterProScan. Compare domain boundaries and architecture with related proteins.

4. Examine structures. Search the PDB for experimental structures and AlphaFold DB for predicted models. Verify that a structure corresponds to the relevant sequence, and inspect experimental details or prediction confidence.

5. Form and qualify a hypothesis. Combine evidence from sequence, domains, and structure to propose a function or testable biological role. Record database versions, identifiers, and analysis settings so the work can be reproduced.

For example, a protein with a statistically strong alignment to a known enzyme and the same conserved catalytic domain has evidence consistent with a related role. The hypothesis is stronger if important residues are conserved and structural evidence places them together in a plausible active site; it still requires appropriate validation.

Reliable interpretation depends on checking evidence, coverage, confidence, and data provenance rather than relying on a single annotation or score.