help with regexpi expression match

I have a question regarding regexpi expression match which may be an easy one (not for me). I have a set of strings from a single cell. An example is as follows:

d2 = {'chromosome 1:NC_011985.1/CP000628.1; chromosome 2:NC_011983.1/CP000629.1; plasmid pAgK84:NC_011994.1/CP000632.1; plasmid pAtK84b:NC_011990.1/CP000630.1; plasmid pAtK84c:NC_011987.1/CP000631.1 chromosome Mycobacterium_bovis_AF2122/97:NC_002945.4/LT708304.1'};

I would like the two names that follow "chromosome" which are found after the ":" and "/" to be picked up with a regexpi expression. so for example, I want to match for NC_011985.1, CP000628.1, NC_011983.1, CP000629.1, NC_002945.4 and LT708304.1 but I want to ignore the other names that follow plasmid. I chose this large string as an example because I wanted the names to be proceeded by the word "chromosome" however as you can see, after the word "chromosome, there may have a number, a word or even nothing, followed by a semicolon ":" a name (that we want to keep) and then another name that follows "/" (that we also want to keep). Keeping all the names in one cell is fine I just want to pick up these names.

if I use the following code:

accession6 = regexpi(d2,'(?<=:)\w+','match');

using this as a base, I do not know how to proceed the match by the word "chromosome" followed by an optional number, or words or even nothing after the word "chromosome" without messing it up . It would have be before the necessary ":" and "/" parts of the expression that go before the name we want to keep.

Any help would be super appreciated.

[Stephen23]

"Any help would be super appreciated"

You might like to download my FEX submission iregexp, an interactive regular expression tool:

https://www.mathworks.com/matlabcentral/fileexchange/48930-interactive-regular-expression-tool

It lets you quickly experiment with different regular expressions and shows all of regexp's outputs in real-time as you type.

Answer 1

[per isakson]

One expression

• 'chromosome' followed by anything up till ':' and one ':'
• capturing group of one or more letter, digit, underscore, and '.' (greedy)
• zero or more of anything up till '/' and one '/'
• capturing group of one or more letter, digit, underscore, and '.' (greedy)

And repeat until no more matches are found

>> cac = regexpi( d2, 'chromosome[^:]+[:]([\w\.]+)[^/]*[/]([\w\.]+)', 'tokens' );
>> cac{:}{:}
ans = 
    'NC_011985.1'    'CP000628.1'
ans = 
    'NC_011983.1'    'CP000629.1'
ans = 
    'NC_002945.4'    'LT708304.1'
>>

If d2 contains one string

>> cac = regexpi( d2{:}, 'chromosome[^:]+[:]([\w\.]+)[^/]*[/]([\w\.]+)', 'tokens' );
>> cac
cac = 
    {1x2 cell}    {1x2 cell}    {1x2 cell}
>>

[Guillaume]

The brackets around the single : and / are not necessary, so the regexp can be slightly simplified to:

regexpi( d2, 'chromosome[^:]+:([\w\.]+)[^/]*/([\w\.]+)', 'tokens' )

I'm not sure I understand the purpose of the [^/]*. As far as I understand it's always going to be a 0 length match.

Another option:

regexpi(d2, 'chromosome[^:;]+:([^/]+)/([^;]+)(?:;|$)', 'tokens')

which matches

• 'chromosome' followed by anything but ':' or ';', then matches ':'
• capture anything but '/', then matches '/'
• capture anything but ';', then matches ';' or the end of the expression

[per isakson]

• "The brackets around the single : and / are not necessary" That's just my way to emphasis that it's not a special character
• "I'm not sure I understand the purpose of the [^/]*" I wasn't sure about the string format. I didn't quite understand the description and though it wouldn't hurt.
• "(?:;|$)" this look-ahead should definitely be added to the end of the expression

[Guillaume]

Small note: (?:;|$) is not a look-anything. It is a non-capturing group.

[per isakson]

Yes, I should know that. However, in this context, I guess they both serve the purpose, i.e. to assert that the name is followed by ';' or the end of the string.

[per isakson]

And an alternative that uses @JM's approach. In a first step match "name slash name" between

• look-behind: (?<=chromosome[^:]+[:])
• look-ahead: (?=;|$)

and in a second step split the two names at slash

cac = regexpi( d2{:}, '(?<=chromosome[^:]+[:])[\w\.]+[/][\w\.]+(?=;|$)', 'match' );
cac = regexp( cac, '/', 'split' );
cac{:}

ans = 
    'NC_011985.1'    'CP000628.1'
ans = 
    'NC_011983.1'    'CP000629.1'
ans = 
    'NC_002945.4'    'LT708304.1'
>>

[per isakson]

chromosome[^:;] with a semi-colon (proposed by @Guillaume) is better than chromosome[^:] without, because the latter will return a plasmid-name-pair if a colon is missing in the string after 'chromosome'. With semi-colon the pair is missed altogether.