thebench.bio #1 Ecosystem for the Bio-Tech Industry
All Sequence Alignment questions

Question 9 of 20 Fundamentals Practical

Which BLAST program would you use for each combination of nucleotide and protein query and database?

Short answer

blastn compares nucleotide with nucleotide and blastp protein with protein. blastx translates a nucleotide query against proteins, tblastn searches a protein against translated nucleotides, and tblastx translates both. For coding sequence, a translated search finds far more distant homologs than blastn.

Why

NCBI's BLAST guide sets out the programs by what is compared. blastn compares nucleotide sequences, with megablast as the default for identifying a sequence or comparing within a species, and discontiguous megablast for comparisons across species. blastp compares proteins, with options such as PSI-BLAST for building a profile of a family. blastx translates a nucleotide query in all reading frames and searches proteins, which is how to find what protein a transcript or contig might encode. tblastn does the reverse, searching translated nucleotide databases with a protein, for example to find an unannotated gene in a new genome. tblastx translates both sides and is the slowest. The choice has a large effect on sensitivity. Pearson notes that protein and translated searches look back 5 to 10 times further in evolution than DNA against DNA.

DNA alignments rarely detect homology beyond 200 to 400 million years, while protein alignments routinely detect it across more than 2.5 billion years. So a coding sequence should usually be searched as protein, with nucleotide searches kept for non-coding sequence or near-identical matches.

How to work it through

  1. Decide whether the query is coding and whether you have it as nucleotide or protein.
  2. Decide what the database holds, protein or nucleotide.
  3. Pick the program from that pairing: blastn, blastp, blastx, tblastn or tblastx.
  4. For nucleotide against nucleotide, choose megablast for near-identical matches and blastn or discontiguous megablast across species.
  5. Check the genetic code setting when translating sequences from mitochondria or unusual organisms.
  6. Read the results by E-value and bit score rather than percent identity.

An example

A team assembles a transcript from a non-model insect and wants to know what it encodes. blastn against nucleotide collections returns nothing below an E-value of 1. blastx against a protein database finds a match to a fruit fly kinase with an E-value of 1e-45 over 300 residues. The translated search detected a homolog that the DNA comparison missed because synonymous changes had erased most nucleotide similarity.

Where the answer stops holding

Translated searches assume a reading frame exists, so they add little for non-coding RNA. Frameshifts from sequencing errors can break a translated alignment into pieces. Very short queries give weak statistics with any program.

A strong answer covers

  • Match the program to query and database types
  • blastx for nucleotide queries against proteins, tblastn the reverse
  • Megablast for near-identical, blastn or discontiguous megablast across species
  • Protein and translated searches reach much deeper in evolution
  • Set the right genetic code for translation

Common mistakes

  • Searching a coding sequence only with blastn and concluding it has no homologs
  • Using megablast to look for distant cross-species matches
  • Forgetting non-standard genetic codes in translated searches