Python Programming for Biology: Bioinformatics and Beyond

Generating an alignment profile

---

The  second  alignment  analysis  we  will  do  is  to  generate  a  profile.  A  profile  is  a  per-

position  statistic  saying  how  much  of  each  kind  of  residue  there  is  at  a  given  location.

Because  a  profile  has  positional  information  it  can  still  be  aligned  just  like  a  simple

sequence alignment, only this time we are dealing with fractions of residue types, rather

than single residues. The Python function to generate a profile from an alignment is given

below.  It  is  very  similar  to  the  consensus  generation,  excepting  that  we  don’t  have  to

choose a winning residue. The profile that is generated at the end is represented as a list of

dictionaries; we locate an alignment position with a list index, then the sub-dictionary for

this  index  gives  the  fractions  of  each  residue  type  present.  Note  how  we  use  the  counts

dictionary initially to do as the name suggests, and store counts, but later it is used to store

the fractions; there is no reason to introduce another variable.

def profile(alignment):

n = len(alignment[0])

nSeq = float(len(alignment))

prof = []

for i in range(n):

counts = {}

for seq in alignment:

letter = seq[i]

if letter == '-':

continue

counts[letter] = counts.get(letter, 0) + 1

for letter in counts:

counts[letter] /= nSeq

prof.append(counts)

return prof

alignment = ['SRPAPVVIILIILCVMAGVIGTILLISYGIRLLIK',

'TVPAPVVIILIILCVMAGIIGTILLISYTIRRLIK',

'HHFSEPEITLIIFGVMAGVIGTILLISYGIRRLIK',

'HEFSELVIALIIFGVMAGVIGTILFISYGSRRLIK']

print(profile(alignment))

# First sub-dict: {'H': 0.5, 'S': 0.25, 'T': 0.25}

Profiles of this kind are often used as position-specific scoring matrices, for example, in

programs  like  PSI-BLAST.

2

 The  idea  here  is  that  you  build  a  profile  for  a  family  of

sequences which have something of interest in common, and then search other sequences

with the whole profile. This allows you to find sequences that share the properties of your

aligned  family  as  a  whole,  rather  than  finding  ones  that  are  similar  to  the  individual

members;  this  increases  the  sensitivity  of  similarity  searches.  A  family  profile  conveys

family-specific  information,  like  the  presence  of  a  highly  conserved  (or  invariant)  site,

which would not be recorded if you looked for sequences with a general substitution table.

---