Mercurial > repos > jay > pdaug_tsvtofasta
comparison PDAUG_Peptide_Ngrams/PDAUG_Peptide_Ngrams.py @ 0:c3f0b3a6339e draft
"planemo upload for repository https://github.com/jaidevjoshi83/pdaug commit a9bd83f6a1afa6338cb6e4358b63ebff5bed155e"
author | jay |
---|---|
date | Wed, 28 Oct 2020 01:47:48 +0000 |
parents | |
children |
comparison
equal
deleted
inserted
replaced
-1:000000000000 | 0:c3f0b3a6339e |
---|---|
1 import matplotlib | |
2 matplotlib.use('Agg') | |
3 import os | |
4 import sys | |
5 sys.path.insert(0, os.path.abspath('..')) | |
6 import quantiprot | |
7 from quantiprot.utils.io import load_fasta_file | |
8 from quantiprot.utils.feature import Feature, FeatureSet | |
9 from quantiprot.metrics.aaindex import get_aa2hydropathy | |
10 from quantiprot.metrics.basic import identity | |
11 from quantiprot.metrics.ngram import pattern_match, pattern_count | |
12 from quantiprot.analysis.ngram import ngram_count | |
13 from quantiprot.analysis.ngram import zipf_law_fit | |
14 from matplotlib import pyplot as plt | |
15 | |
16 | |
17 def Run_ngrams(fasta1, fasta2, OutFile ): | |
18 | |
19 alphasyn_seq = load_fasta_file(fasta1) | |
20 amyload_pos_seq = load_fasta_file(fasta2) | |
21 | |
22 fs_aa = FeatureSet("aa patterns") | |
23 fs_aa.add(identity) | |
24 fs_aa.add(pattern_match, pattern='VT', padded=True) | |
25 fs_aa.add(pattern_count, pattern='VT') | |
26 | |
27 result_seq = fs_aa(alphasyn_seq) | |
28 | |
29 fs_hp = FeatureSet("hydropathy patterns") | |
30 fs_hp.add(Feature(get_aa2hydropathy())) | |
31 fs_hp.add(Feature(get_aa2hydropathy()).then(pattern_match, pattern=[0.0, 2.0], | |
32 metric='taxi', radius=1.0)) | |
33 result_seq2 = fs_hp(alphasyn_seq) | |
34 result_freq = ngram_count(alphasyn_seq, n=2) | |
35 result_fit = zipf_law_fit(amyload_pos_seq, n=3, verbose=True) | |
36 | |
37 counts = sorted(result_fit["ngram_counts"], reverse=True) | |
38 ranks = range(1, len(counts)+1) | |
39 | |
40 slope = result_fit["slope"] | |
41 harmonic_num = sum([rank**-slope for rank in ranks]) | |
42 fitted_counts = [(rank**-slope) / harmonic_num * sum(counts) for rank in ranks] | |
43 | |
44 plt.plot(ranks, counts, 'k', label="empirical") | |
45 plt.plot(ranks, fitted_counts, 'k--', | |
46 label="Zipf's law\nslope: {:.2f}".format((slope))) | |
47 plt.xlabel('rank') | |
48 plt.ylabel('count') | |
49 plt.xscale('log') | |
50 plt.yscale('log') | |
51 plt.legend() | |
52 | |
53 plt.savefig(OutFile) | |
54 | |
55 if __name__=="__main__": | |
56 | |
57 | |
58 import argparse | |
59 | |
60 parser = argparse.ArgumentParser() | |
61 | |
62 parser.add_argument("-f1", "--Fasta1", | |
63 required=True, | |
64 default=None, | |
65 help="First fasta file") | |
66 | |
67 parser.add_argument("-f2", "--Fasta2", | |
68 required=True, | |
69 default=None, | |
70 help="Second fasta file") | |
71 | |
72 | |
73 parser.add_argument("--OutFile", | |
74 required=True, | |
75 help="HTML out file", | |
76 default="report.html") | |
77 | |
78 | |
79 args = parser.parse_args() | |
80 | |
81 Run_ngrams(args.Fasta1, args.Fasta2, args.OutFile) | |
82 |