CODEWITHSUNDEEP
pythonlistdictionarycounter
biogeek
biogeek

Reputation: 197

How to extract and divide values from dictionary with another in Python?

sample = [['CGG','ATT'],['GCGC','TAAA']]

#Frequencies of each base in the pair
d1 = [[{'G': 0.66, 'C': 0.33}, {'A': 0.33, 'T': 0.66}], [{'G': 0.5, 'C': 0.5}, {'A': 0.75, 'T': 0.25}]]

#Frequencies of each pair occurring together

d2 = [{('C', 'A'): 0.33, ('G', 'T'): 0.66}, {('G', 'T'): 0.25, ('C', 'A'): 0.5, ('G', 'A'): 0.25}]

The Problem:

Consider the first pair : ['CGG','ATT']

How to calculate a, where a is :

float(a) = (freq of pairs) - ((freq of C in CGG) * (freq of A in ATT))

eg. in CA pairs, float (a) = (freq of CA pairs) - ((freq of C in CGG) * (freq of A in ATT))

Output a = (0.33) - ((0.33) * (0.33)) = 0.222222

Calculating "a" for any one combination (either CA pair or GT pair)

Final Output for sample : a = [0.2222, - 0.125]

How to calculate b, where b is :

float (b) = (float(a)^2)/ (freq of C in CGG) * (freq G in CGG) * (freq A in ATT) * (freq of T in ATT)

Output b = 1

Do this for the entire list

Final Output for sample : b = [1, 0.3333]

I do not know how to extract the required values from d1 and d2 and perform the mathematical operations.

I tried to write the following code for value of a

float a = {k: float(d1[k][0]) - d2[k][0] * d2[k][1]for k in d1.viewkeys() & d2.viewkeys()}

But, it does not work. Also, I prefer a for loop instead of comprehensions

My attempt to write (a pretty flawed) for-loop for the above:

float_a = []
for pair,i in enumerate(d2):
    for base,j in enumerate(d1):
        float (a) = pair[i][0] - base[j][] * base[j+1][]
        float_a.append(a)

float_b = []
  for floata in enumerate(float_a):
    for base,j in enumerate(d1):
        float (b) = (float(a) * float(a)) - (base[j] *    base[j+1]*base[j+2]*base[j+3])
        float_b.append(b)

Upvotes: 4

Views: 348

Answers (1)

CodeSurgeon
CodeSurgeon

Reputation: 2465

Usually when there is a tricky problem like this with multiple formulas and intermediate steps, I like to modularize it by splitting the work into several functions. Here is the resulting commented code which handles the cases in the original question and in the comments:

from collections import Counter

def get_base_freq(seq):
    """
    Returns the normalized frequency of each base in a given sequence as a dictionary.
    A dictionary comprehension converts the Counter object into a "normalized" dictionary.
    """
    seq_len = len(seq)
    base_counts = Counter(seq)
    base_freqs = {base: float(count)/seq_len for base, count in base_counts.items()}
    return base_freqs

def get_pair_freq(seq1, seq2):
    """
    Uses zip to merge two sequence strings together.
    Then performs same counting and normalization as in get_base_freq.
    """
    seq_len = len(seq1)
    pair_counts = Counter(zip(seq1, seq2))
    pair_freqs = {pair: float(count)/seq_len for pair, count in pair_counts.items()}
    return pair_freqs

def calc_a(d1, d2):
    """
    Arbitrarily takes the first pair in d2 and calculates the a-value from it.
    """
    first_pair, pair_freq = d2.items()[0]
    base1, base2 = first_pair
    a = pair_freq - (d1[0][base1]*d1[1][base2])
    return a

def calc_b(a, d1):
    """
    For this calculation, we need to use all of the values from d1 and multiply them together.
    This is done by merging the two sequence half-results together and multiplying in a for loop.
    """
    denom_ACGT = d1[0].values() + d1[1].values()
    denom = 1
    for val in denom_ACGT:
        denom *= val
    b = a*a/float(denom)
    return b

if __name__ == "__main__":
    sample = [['CGG','ATT'], ['GCGC','TAAA'], ['ACAA','CAAC']]
    b_result = []
    for seq_pair in sample:
        d1 = [get_base_freq(seq) for seq in seq_pair]
        d2 = get_pair_freq(*seq_pair)
        a = calc_a(d1, d2)
        b = calc_b(a, d1)
        b_result.append(b)
    print b_result

Let me know if anything needs to be cleared up or if it fails for a case which I have not considered!

Upvotes: 1

Related Questions