def overlap(self):
pdb_uni = self.read_pdb_uni()
cb_ids, cb_seqs = tools_fasta.fasta_to_id_seq(self.bc90)
cb_pdb_unis = {}
cb_pdbs = []
for id in cb_ids:
if id in pdb_uni:
cb_pdb_unis[pdb_uni[id]] = id
cb_pdbs.append(pdb_uni[id])
pdb_ids, pdb_seqs = tools_fasta.fasta_to_id_seq(self.pdb90)
print("Proteins overlapping between the PDB and BC datasets")
for pdb_id in pdb_ids:
if pdb_id in cb_pdbs:
print(pdb_id)
print(cb_pdb_unis[pdb_id])
def get_human(self):
pids, seqs = tools_fasta.fasta_to_id_seq(self.human_fp)
ns = NormScore()
scores = ns.lc_norm_score(seqs)
proteome = ['Human'] * len(pids)
org = ['Human'] * len(pids)
return pids, proteome, org, scores
def get_yeast(self):
pids, seqs = tools_fasta.fasta_to_id_seq(self.yeast_fp)
ns = NormScore()
scores = ns.lc_norm_score(seqs)
proteome = ['Yeast'] * len(pids)
org = ['Yeast'] * len(pids)
return pids, proteome, org, scores
def train_df(self):
pdb_pids, pdb_seqs = tools_fasta.fasta_to_id_seq(self.pdb_fpi)
pdb_lens = tools_fasta.get_lengths(pdb_seqs)
bc_pids, bc_seqs = tools_fasta.fasta_to_id_seq(self.bc_fpi)
bc_lens = tools_fasta.get_lengths(bc_seqs)
lens = bc_lens + pdb_lens
pids = bc_pids + pdb_pids
seqs = bc_seqs + pdb_seqs
y = [0] * len(bc_pids) + [1] * len(pdb_pids)
df_dict = {
'Protein ID': pids,
'Sequence': seqs,
'Length': lens,
'y': y
}
cols = ['Protein ID', 'y', 'Sequence', 'Length']
df = pd.DataFrame(df_dict, columns=cols)
df.to_csv(self.train_fpo, sep='\t')
def write_scores(self):
ids, seqs = tools_fasta.fasta_to_id_seq(self.all_fpi)
ns = NormScore()
scores = ns.lc_norm_score(seqs)
df_out = pd.DataFrame({
'Protein ID': ids,
'LC Score': scores
},
columns=['Protein ID', 'LC Score'])
df_out = df_out.sort_values(by='LC Score', ascending=False)
print(df_out)
df_out.to_csv(self.all_fpo, sep='\t')
def get_scores(self):
pbodies = self.get_pbody()
pids, seqs = tools_fasta.fasta_to_id_seq(self.yeast_fasta)
pseqs = []
ppids = []
for pid, seq in zip(pids, seqs):
if pid in pbodies:
pseqs.append(seq)
ppids.append(pid)
ns = NormScore()
scores = ns.lc_norm_score(pseqs)
df_dict = {'Protein ID': ppids, 'LC Score': scores}
df_out = pd.DataFrame(df_dict)
df_out.to_csv(self.yeast_scores, sep='\t')
def read_fasta(self):
pids, seqs = tools_fasta.fasta_to_id_seq(self.fasta_fpi)
norm = NormScore()
# ent1[211:457]
ent1 = seqs[0]
#print(ent1[211:457])
ent1wo = ent1[:211] + ent1[457:]
#print(norm.lc_norm_score([ent1wo]))
#print(norm.lc_norm_score([ent1]))
# ent2[224:616]
ent2 = seqs[1]
#print(ent2[224:616])
ent2wo = ent2[:224] + ent2[616:]
#print(norm.lc_norm_score([ent2wo]))
# yap1801[351:638]
yap1801 = seqs[2]
#print(yap1801[351:638])
yap1801wo = yap1801[:351] + yap1801[638:]
#print()
#print(norm.lc_norm_score([yap1801]))
#print(norm.lc_norm_score([yap1801wo]))
# yap1802[319:569]
yap1802 = seqs[3]
#print(yap1802[319:569])
yap1802wo = yap1802[:319] + yap1802[569:]
#print(norm.lc_norm_score([yap1802wo]))
# sla1[954:1244]
sla1 = seqs[4]
print(len(sla1))
print(sla1[954:1244])
print()
ns = tools_lc.display_lc(sla1, self.k, self.lca, self.lce)
print(sla1)
print(ns)
sla1wo = sla1[:954] + sla1[1244:]
print(norm.lc_norm_score([sla1wo]))
print(norm.lc_norm_score([sla1]))
#sla2[348:442]
sla2 = seqs[5]
#print(sla2[348:442])
sla2wo = sla2[:348] + sla2[442:]
#print(norm.lc_norm_score([sla2wo]))
#print(norm.lc_norm_score([sla2]))
# sup35[0:123]
sup35 = seqs[6]
def run(self):
ids, seqs = tools_fasta.fasta_to_id_seq(self.puncta)
df = pd.read_csv(self.pfam_puncta, sep='\t', index_col=0)
new_seqs = []
below = 0
above = 0
norm_scores = []
fl_norm_scores = []
for id, seq in zip(ids, seqs):
ndf = df[df['uniprot_acc'] == id]
ndf = ndf.sort_values(by='seq_start')
segmented = self.segment_seq(seq, ndf)
total = 0
for item in segmented:
total += len(item)
if total >= 100:
above += 1
fl_score, fl_length = self.get_segment_scores([seq])
fl_norm = self.norm_function([fl_score], [fl_length])
raw_score, length = self.get_segment_scores(segmented)
norm_score = self.norm_function([raw_score], [length])
norm_scores.append(norm_score[0])
fl_norm_scores.append(fl_norm[0])
else:
below += 1
print(above)
print(below)
print(np.mean(norm_scores))
print(np.mean(fl_norm_scores))
print(np.median(norm_scores))
print(np.median(fl_norm_scores))
plt.hist(fl_norm_scores,
alpha=0.5,
bins=20,
range=(-100, 200),
label='Full length scores')
plt.hist(norm_scores,
alpha=0.5,
bins=20,
range=(-100, 200),
label='Outside Pfam scores')
plt.legend()
plt.show()
def concat_train(self):
bc_pids, bc_seqs = tools_fasta.fasta_to_id_seq(self.bc_fpi)
bc_lens = tools_fasta.get_lengths(bc_seqs)
pdb_df = pd.read_csv(self.pdb_fpi, sep='\t', index_col=0)
pdb_pids = list(pdb_df['Protein ID'])
pdb_seqs = list(pdb_df['Sequence'])
pdb_lens = list(pdb_df['Length'])
pids = bc_pids + pdb_pids
seqs = bc_seqs + pdb_seqs
lens = bc_lens + pdb_lens
y = [0] * len(bc_pids) + [1] * len(pdb_pids)
cols = ['Protein ID', 'y', 'Length', 'Sequence']
df_dict = {
'Protein ID': pids,
'Sequence': seqs,
'Length': lens,
'y': y
}
df = pd.DataFrame(df_dict, columns=cols)
df.to_csv(self.fpo, sep='\t')
def with_pfam(self, fasta_fp, pfam_fp, fpo):
"""
How many proteins in the set have pfam domains?
What is the fraction occupied by pfam domains?"""
df = pd.read_csv(pfam_fp, sep='\t')
pfam_ids = list(set(df['uniprot_acc']))
pids, seqs = tools_fasta.fasta_to_id_seq(fasta_fp)
print(len(pids))
nopfam_ids = list(set(pids) - set(pfam_ids))
nopfam_seqs = []
for pid, seq in zip(pids, seqs):
if pid in nopfam_ids:
nopfam_seqs.append(seq)
ns = NormScore()
scores = ns.lc_norm_score(nopfam_seqs)
df_out = pd.DataFrame({
'UniProt ID': nopfam_ids,
'LC Score': scores
},
columns=['UniProt ID', 'LC Score'])
df_out = df_out.sort_values(by='LC Score', ascending=False)
df_out.to_csv(fpo, sep='\t')
def percent_pfam(self, fasta_fp, pfam_fp, fpo):
df = pd.read_csv(pfam_fp, sep='\t')
pids, seqs = tools_fasta.fasta_to_id_seq(fasta_fp)
frac_pfam = []
for id, seq in zip(pids, seqs):
ndf = df[df['uniprot_acc'] == id]
ndf = ndf.sort_values(by='seq_start')
segmented = self.segment_seq(seq, ndf)
len_seg = 0
for seg in segmented:
len_seg += len(seg)
frac_pfam.append(float(len(seq) - len_seg) / float(len(seq)))
ns = NormScore()
scores = ns.lc_norm_score(seqs)
df_out = pd.DataFrame(
{
'Uniprot ID': pids,
'LC Score': scores,
'Pfam Fraction': frac_pfam
},
columns=['Uniprot ID', 'LC Score', 'Pfam Fraction'])
df_out = df_out.sort_values(by='LC Score', ascending=False)
df_out.to_csv(fpo, sep='\t')
print(np.mean(frac_pfam))
def get_pids(self, fasta):
pids, seqs = tools_fasta.fasta_to_id_seq(fasta)
return pids
