The fasta parsing rule in XiSearch

[WeiqiangChen]

XiSearch parses fasta header differently than other software, such as FragPipe, pLink, and Spectronaut.

When I create a custom fasta sequence, I set the header to the following:

xx|UniqueIdentifier|UniqueIdentifier_SYNTH ProteinName OS=Synthetic OX=0000 GN=GeneName PE=1 SV=1

XiSearch will parse the ID to "xx" and Accession to "xx". And the resulting CSM table is a mess if there are multiple custom fasta sequences. Also, https://xiview.org/ goes crazy when trying to expand the target protein.

Only when I change the header to

sp|UniqueIdentifier|UniqueIdentifier_SYNTH ProteinName OS=Synthetic OX=0000 GN=GeneName PE=1 SV=1

Xisearch works normally.

Is it possible to change the parsing rule so that I do not have to change the fasta header for XiSearch?

Thank you!
WeiQiang

[lutzfischer]

xisearch "recognises" several flavours of fasta header

XiSearch/src/main/java/rappsilber/ms/sequence/fasta/FastaHeader.java

Lines 38 to 45 in 162cd77

	Pattern.compile("^([\\w-]+)\\s+(?:\\w+)\\sprotein"), // SGD Header
	Pattern.compile("^(?:gi\\|(\\w+)\\|.*)"), // GI header
	//Pattern.compile("^\\w+\\:(\\w+)\\W+(?:(?:.*)Gene_Symbol=(?:[^\\s]*))*(.*)"), // IPI
	Pattern.compile("^(?:REV_|rev_)?sp\\|([\\w-]+)\\|(?:.*)"), // sprot version 1
	Pattern.compile("^(?:REV_|rev_)?tr\\|([\\w-]+)\\|(?:.*)"), // sprot version 1
	Pattern.compile("^(\\w+)\\|(?:.*)"), //sprot version 2
	Pattern.compile("^([^\\s]*)\\s+[^\"]*\"(?:.*)\""), //sprot version 2
	Pattern.compile("^([\\w\\.]+)\\|(([^\\|]*)\\|([^\\|]*)\\|([^\\|]*))(\\|.*)?"), //sanger

I think your specific one hits a "known" format and currently there is no way to overwrite that. If it would not be recognised then the

default_accession_re:^\s*([^\s]+)(?:\s+.)
default_name_re:^$
default_description_re:^\s(.[^\s])\s

settings would be used to define how to parse the fasta headers. These can be changed in the config (can be defined in custom setting in the GUI)

Source for that behaviour is, that users frequently use fasta files with mixed flavours of fasta headers.

[grandrea]

You can define your own parsing rules in your config file

https://github.com/Rappsilber-Laboratory/XiSearch/tree/162cd77b2e36c9f58119a6bca7a673ef1350103e?tab=readme-ov-file#the-files-tab

default_accession_re:	custom regular expression to define protein accession
default_name_re:	custom regular expression to define protein name
default_description_re:	custom regular expression to define protein description

[WeiqiangChen]

Do you mean to add the following lines to the .conf file?

default_accession_re:^[^|]|([^|]+)|
default_name_re:^[^|]|[^|]+|([^ ]+)
default_description_re:^[^|]*|[^|]+|[^ ]+ (.+?)(?= OS=)

Even if this makes XiSearch work, I guess XiView website will still go crazy.

[lutzfischer]

If you upload mzIdentML, what ever xiSEARCH assumes to be the accession, name and description will be used in xiVIEW. It basically ignores the fasta and loads all the informatin directly from the mzIdentML file.
However I think at the moment xiSEARCH would not use the default_*_re settings as it "recognises" the format with "^(\w+)|(?:.*)" and that then takes precedence before the default regular expressions.

To change that I would need to add optional fixed parsing system.

Can do but not this week anymore.

[WeiqiangChen]

Thanks..
Maybe just add a swissprot-like. rule. to be any 2 character string than sp or tr....

[grandrea]

Do you mean to add the following lines to the .conf file?

default_accession_re:^[^|]|([^|]+)| default_name_re:^[^|]|[^|]+|([^ ]+) default_description_re:^[^|]*|[^|]+|[^ ]+ (.+?)(?= OS=)

Even if this makes XiSearch work, I guess XiView website will still go crazy.

This should be wholly compatible with xiview once you upload mzIdentML. name, accession and description get passed to xiFDR mzIdentML writer and then parsed by xiview.