makeblastdb Error: ncbi::objects::CSeq_id::x_Init() - Unsupported ID type 1196094

Question

[chenyil91]

Hi everyone! Recently, I run makeblastdb on the server . the order is " makeblastdb -in kraken2_blast -dbtype nucl -out nt -title "nt" -parse_seqids -hash_index"; but there is an error:

file: /home/lcy/winterbee/database/kraken-blast/nt.nsq
file: /home/lcy/winterbee/database/kraken-blast/nt.nsi
file: /home/lcy/winterbee/database/kraken-blast/nt.nsd
file: /home/lcy/winterbee/database/kraken-blast/nt.nhi
file: /home/lcy/winterbee/database/kraken-blast/nt.nhd
file: /home/lcy/winterbee/database/kraken-blast/nt.nog

Error: NCBI C++ Exception:
    T0 "/home/coremake/release_build/build/PrepareRelease_Linux64-Centos_JSID_01_350334_130.14.22.10_9008__PrepareRelease_Linux64-Centos_1481139955/c++/compilers/unix/../../src/objects/seq/../seqloc/Seq_id.cpp", line 1911: Error: ncbi::objects::CSeq_id::x_Init() - Unsupported ID type 1196094

Dose anyone know how to solve the problem? Looking forward to your reply~ Thank you! Chenyi

[lieven.sterck]

Looks like there might be a problem with your input file (headers). Can you post the output of grep '>' kraken2_blast | head . Does this number/ID 1196094 appears in your fasta input file?

Answer 1

[Michael]

Blast is rather picky about identifiers. As soon as you have identifiers containing pipes |, it assumes your are trying to give it NCBI-style identifiers. But the behavior of makeblastdb may have changed slightly, so I tested a few options. Also, makeblastdb will only report the first error it encounters.

>1196094|1196094|1196094|1196094 : this works fine, all numeric doesn't seem to trigger NCBI-schema parsing
ACGTTTT
>gi|1196094|blubb|1196094 : this doesn't work, blubb is an unsupported ID type 
ACGTTTT
>jgi|Naegr1|60034|estExt_fgeneshHS_pg.C_890017 : This is a JGI id, just take it! Bummer!
TTTTTTTTT
>mjgi|Naegr1|60034|estExt_fgeneshHS_pg.C_890017 : This works, first entry length > 3 doesn't seem to trigger NCBI-schema parsing
TTTTTTTTT

>bla|buzz|foob|1196094 : bla (and foo, bar) is obviously not a recognized ID type
ACGT
>lcl|1|1196094|2 : This likely how the header was formatted in OP's file, a valid ID type, followed by an invalid one: Unsupported ID type 1196094
ACGTTTT
>lcl|1|lcl|2|1196094 : Or like this, 2 valid pairs, plus an additional number
ACGTTTT
>lcl|1|lcl|2|lcl|3|1196094 : Or like this, you name it
ACGTTTT

lcl (local) is a valid id type, so if you want to fix the file, while keeping the number, add lcl| right after the last |, such that the header looks like:

>lcl|1|lcl|1196094

Or, replace all | with another character, or run without -parse_seqids, or ...

The following blog may also be related: https://blastedbio.blogspot.com/2012/10/my-ids-not-good-enough-for-ncbi-blast.html

[GenoMax]

See: https://www.ncbi.nlm.nih.gov/books/NBK569841/

The identifier should begin right after the “>” sign on the definition line and contain no spaces and the -parse_seqids flag should be used. In general, you should not use a “|” (bar) in your identifier. The “|” (bar) is a reserved character for the NCBI FASTA ID parser and makeblastdb will return an error unless the bar is used in a specific manner described at https://ncbi.github.io/cxx-toolkit/pages/ch_demo#ch_demo.T5

[Michael]

Sure, the idea was to find out what to do as a quick fix, if you got a file that is ill-formatted (like JGI file I have in the example) or the kraken_blast file OP got. I was just trying this with a JGI protein annotation file and needed to keep the identifiers intact.