Sequence Analysis
Analyze protein sequences for homology, structure prediction, and functional annotation to support target identification and validation.
Sequence Analysis
Sequence analysis is the foundation of target identification in drug discovery. By examining a protein's amino acid sequence, you can infer evolutionary relationships, predict three-dimensional structure, identify functional domains, and assess druggability. This workflow chains sequence similarity search, structure prediction, and database annotation into a comprehensive target characterization pipeline.
Overview
Starting from a protein sequence of interest -- whether from a genome-wide association study hit, a differential expression experiment, or a literature lead -- this workflow uses BLAST to find homologous proteins across species, ESMFold to predict the three-dimensional structure, and UniProt to retrieve curated functional annotations, disease associations, and post-translational modification data. Together, these tools transform a raw amino acid sequence into an actionable target profile.
Tool Chain
- BLAST -- Sequence similarity search against curated databases
- ESMFold -- AI-powered protein structure prediction
- UniProt -- Protein sequence and functional annotation database
Prerequisites
- A protein sequence in FASTA format or a UniProt/NCBI accession number
- Basic understanding of sequence alignment metrics (E-value, percent identity)
- Clarity on the biological question driving the analysis
Step-by-Step Guide
Retrieve or prepare the sequence
If you have a UniProt accession or gene name, use the Biotite sequence fetch tool to retrieve the canonical protein sequence with metadata. If starting from a nucleotide sequence, translate to amino acids first. Verify the sequence contains only standard amino acid characters and is in the correct reading frame.
Search for homologs with BLAST
Submit the sequence to BLAST using blastp (protein-protein) against the non-redundant (nr) database for broadest coverage, or against UniProt/SwissProt for curated hits only. Set the E-value threshold to 1e-5 for initial searches (relaxing to 1e-3 for distant homology detection). Review the top hits, paying attention to percent identity, query coverage, and taxonomic distribution. Homologs across multiple species suggest conserved function and validated druggability.
Analyze sequence conservation
Examine the BLAST alignments for conserved regions that likely correspond to functional domains, catalytic residues, or binding sites. Highly conserved positions across distantly related species are strong candidates for functional importance. Note any gaps or insertions that may indicate species-specific adaptations or isoform differences.
Predict three-dimensional structure
Submit the sequence to ESMFold for structure prediction. Review the resulting 3D model with pLDDT confidence coloring. Regions with pLDDT above 90 (dark blue/green) represent high-confidence structural elements suitable for structure-based drug design. Regions below 50 (red/orange) may be intrinsically disordered or poorly predicted and should be interpreted cautiously.
Annotate function from UniProt
Query UniProt with the protein accession to retrieve curated annotations including gene ontology (GO) terms, enzyme classification, subcellular localization, tissue expression, disease associations, known ligands and inhibitors, and post-translational modifications. Cross-reference the UniProt domain annotations with the structural model from ESMFold to map functional sites onto the three-dimensional structure.
Assess druggability
Combine the sequence analysis results into a druggability assessment. Key indicators include the presence of well-defined binding pockets (from ESMFold structure), known small-molecule interactions (from UniProt and homolog data), conservation of the binding site across species (from BLAST analysis), and disease association strength. Proteins with deep, hydrophobic pockets and validated genetic associations make the strongest drug targets.
Document and export
Compile the target characterization into a structured report: sequence metadata, BLAST homology results, ESMFold structure with confidence assessment, UniProt annotations, and druggability evaluation. Export the structure as PDB for downstream docking workflows.
Expected Outputs
- BLAST alignment results with homolog lists, E-values, and percent identity
- ESMFold 3D structure with per-residue pLDDT confidence scores
- UniProt annotation summary (GO terms, disease associations, PTMs, ligands)
- Domain architecture map aligned to sequence positions
- Druggability assessment with supporting evidence
- PDB structure file for downstream structure-based workflows
Tips and Best Practices
- Search against both nr and SwissProt databases. SwissProt provides higher-quality curated annotations, while nr captures more distant homologs and recently deposited sequences.
- For targets with no close homologs (percent identity below 30%), consider using PSI-BLAST or HHblits for more sensitive remote homology detection.
- Cross-validate ESMFold predictions against any available experimental structures from PDB. If the RMSD between predicted and experimental structures is below 2 Angstroms, the model is reliable for docking.
- Pay attention to isoform differences in UniProt. The canonical sequence may not match the disease-relevant isoform expressed in your tissue of interest.
- Use the sequence conservation analysis to guide mutagenesis experiments: highly conserved residues are likely essential for function and make poor sites for affinity tags or truncation constructs.
- For multi-domain proteins, consider folding individual domains separately with ESMFold for better prediction accuracy, then assemble the full-length model.
Was this page helpful?