liveBook · Manning

Multiple sequence alignment

As explained in Build AI Drug Discovery Pipelines

A multiple sequence alignment is an arrangement of related biological sequences in aligned positions to reveal conserved and correlated patterns.

A multiple sequence alignment (MSA) lines up related protein sequences so conserved positions and correlated mutations become visible. It begins with one FASTA sequence and a database search for homologous proteins. section 11.3 section 11.3.1

In protein-structure prediction, an MSA supplies evolutionary signal when experimental structures are unavailable. Correlations between positions can indicate physical contacts: compensatory mutations may preserve structure or function when interacting residues change. section 11.3.1

A practical preprocessing workflow is:

  • Search UniClust30 with HHblits to generate an alignment from a protein FASTA file. section 11.3
  • Filter sequences with HHfilter to remove sequences exceeding 90% identity, reducing redundancy and computational cost. section 11.3
  • Pass the filtered alignment to Pconsc4 for contact-map prediction and use the MSA to calculate a position-specific scoring matrix (PSSM). section 11.3 section 11.3.1

Command-line preprocessing

hhblits -i protein.fasta -d uniclust30 -o protein.aln -n 3 hhfilter -i protein.aln -o protein.filtered.aln -id 90 The output is a filtered MSA, not a contact map. It becomes a prerequisite for downstream contact-map and PSSM computation. section 11.3

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Explains retrieval, redundancy filtering, evolutionary coupling, and downstream uses, but stops short of detailing alignment algorithms.
Examples
Demonstrates HHblits and HHfilter commands and a pipeline branching toward Pconsc4 contact maps and PSSM features.
What it assumes
Assumes a protein FASTA sequence and access to sequence databases and external bioinformatics tools.
What you can do after
Operate/configure HHblits and HHfilter to produce a filtered alignment for downstream protein feature generation.
How it connects
Connects homologous proteins, evolutionary coupling, contact maps, PSSMs, protein graphs, and graph neural networks.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.