replacing part of a string with value from another column postgresql

Question

I have a table with information of mutations in a column is the amino acid change in three letters code as follow:

Amino acid change
------------------------ 
NP_006209.2:p.1025
NP_203524.1:p.12  
NP_000537.3:p.273 
NP_004324.2:p.600 
NP_000537.3:p.215

In another table I have the three letters code and the one letter code of the amino acids as follow:

three_letters|one_letters
-------------|-----------
Ala          |A          
Arg          |R          
Asn          |N          
Asp          |D             
...
Val          |V          
Asx          |B          
Glx          |Z          
Ter          |*          

I need a new column in my table of mutations with the amino acids in one letter code as follow:

new column
----------- 
p.1025  
p.12    
p.273   
p.600   
p.215

Answer 1

You can solve this using a regular expression so long as the change code is always three letters followed by one or more digits followed by three letters.

regexp_match(change, 'p.(\D{3})(\d+)(\D{3})')

That returns an array that can be used to join to your lookup table and then reconstruct the shortened code.

with split as (
  select *, 
         regexp_match(change, 'p.(\D{3})(\d+)(\D{3})') as parts
    from changes
)
select s.*, 
       concat('p.', 
              coalesce(x1.one_letters, '?'), 
              parts[2], 
              coalesce(x2.one_letters, '?') 
       ) as encoded_change
  from split s
  left join xlate x1 on x1.three_letters = s.parts[1]
  left join xlate x2 on x2.three_letters = s.parts[3];

Answer 2

An alternative to the solution proposed by @Mike Organek is to create a short function to do this conversion for you.

Data Sample:

CREATE TEMPORARY TABLE map (three_letters text, one_letters text);
INSERT INTO map 
VALUES ('Val','V'),('Glu','E'),('Thr','T'),('Ala','A');

Function:

CREATE OR REPLACE FUNCTION change_amino_acid(text)
RETURNS TEXT AS $BODY$
DECLARE i RECORD; acid TEXT;
BEGIN
acid := trim((string_to_array($1, ':p.'))[2]);
FOR i IN SELECT * FROM map 
  WHERE three_letters = ANY(regexp_split_to_array(acid, '\d+'))
LOOP
  acid := replace(acid,i.three_letters,i.one_letters);
END LOOP;
RETURN 'p.'||acid;
END; $BODY$ LANGUAGE plpgsql;

How to call the function:

SELECT 
  change_amino_acid('NP_006209.2:p.Thr1025Ala'),
  change_amino_acid('NP_004324.2:p.Val600Glu');

 change_amino_acid | change_amino_acid 
-------------------+-------------------
 p.T1025A          | p.V600E

After that all you need to do is to UPDATE your table using the function

UPDATE my_table 
SET newcolum = change_amino_acid(long_amino_acid);

Answer 3

Your string is in a very particular format. The prefix looks like a fixed length. Then it is followed by three characters, a number (presumably a position), and then three more characters.

If this is always the case, you don't need any real sophisticated machinery for the replacement. You can just use string operations:

with replacements as (
      select 'Thr' as three_letters, 'T' as one_letter union all
      select 'Ala' as three_letters, 'A' as one_letter
     ) 
select v.*, 
       left(mutation, 14) || r1.one_letter || replace(substr(mutation, 18), r2.three_letters, r2.one_letter)
from (values ('NP_006209.2:p.Thr1025Ala')) v(mutation) left join
     replacements r1
     on r1.three_letters = substr(mutation, 15, 3) left join
     replacements r2
     on r2.three_letters = right(mutation, 3);

I would actually recommend that you change the data structure so the values are not all encoded in a single string. Put the results in multiple columns:

• name
• from_amino_acid
• to_amino_acid
• position

Actually, I don't know what is happening before the : , nor whether the p. is important. You might want to split that into more than one column as well. You can use logic like this to split the string:

select split_part(mutation, ':', 1) as name,
       substring(split_part(mutation, ':', 2), 3, 3) as from_amino_acid,
       (regexp_matches(split_part(mutation, ':', 2), '[0-9]+'))[1] as position,
       right(mutation, 3) as to_amino_acid
from (values ('NP_006209.2:p.Thr1025Ala')) v(mutation);

This would simplify your SQL and probably your analyses as well.