Tool that builds multi-fasta/seqinfo/taxtable from query sequence and its blast results

[tyleraland]

This tool's development is primarily motivated by the CAP requirement to validate our primer sequences'. This tool would also enable a future tool for building de novo trees from selected sequences.

Inputs: a single bait sequence (named, inside a fasta file).

Outputs:

• A multi-fasta file of hits for that bait, containing sequences of variable length.
• Seq info (csv) file, with one line for each sequence in the multi-fasta.
• Tax table, with taxonomic information for each sequence in the multi fasta.

What it does:

• BLAST using the query bait sequence
• Using the blast results (metadata), download the corresponding sequences and store to file

Questions:

• Blasting against NCBI will be the first target. Will we want to blast against other databases?
• Do we want to require minimum/maximum sequence lengths or coverage? Are reference genomes too big? Are short fragments too small?
• Do we want to judge sequence quality (e.g. published vs. not) before or after we download?
• Is this just an extension to bioy blast?

Related software:

• https://github.com/fhcrc/deenurp ... builds sequence reference packages
• https://bitbucket.org/uwlabmed/blast2tree ... from query sequence, return blast result sequences

Future work (once this tool is built):

• Tool to crop the above unaligned multi-fasta

[nhoffman]

If you want to provide tax_ids at a uniform rank in the seq_info file, you will need to use taxtastic and include a taxonomy database as one of the inputs.

[tyleraland]

Re-envisioning the path forward.

Rather than re-implement a tool to perform/wrap BLAST, this proposed tool will use the output of (bioy) BLAST to download sequences. In particular, sseqids uniquely specify sequences to pull down from NCBI.

(Revised) Inputs: a file of unique identifiers (sseqid's / gi's), each corresponding to a sequence to download from NCBI.

It's conceivable that we might want to extend this feature to arbitrary (local/URI) sequence databases (that is, provide a sseqid and a database and download their sequences) but it's not trivial to reconstruct a sequence from, say, a BLAST database. What we could do is just search a fasta file for a gi and spit the sequence out. Perhaps we can add this functionality in the future, but in the mean time this can be done with seqmagick.

For the immediate future we should target NCB.

[tyleraland]

Needs satisfied by ...

• Extending bioy blast to perform remote (NCBI) blast
• ncbi_fetch subcommand to pull sequences from NCBI by id