def filter_multi_orths(arr_in, basesp, cutoff):
"""
For every interaction without base basespecies evidence, remove other basespecies
evidence for that interaction when the base basespecies side of the orthogroup
is greater than 1.
"""
print "Filtering: require rows w/o %s > %s to have single orths" % (basesp,
cutoff)
arr = ut.arr_copy(arr_in)
basesp_cols = [n for n in arr.dtype.names[3:] if n[:2]==basesp]
assert len(basesp_cols)>0, 'No base species data.'
maxes = arr_collist_maxes(arr, [basesp_cols])
othersps = species_list(arr.dtype.names[3:])
othersps.remove(basesp)
spcols = [(sp, [n for n in arr.dtype.names[3:] if n[:2]==sp])
for sp in othersps]
ogs_all = orth.all_ogroup_sizes(basesp, othersps)
cleared = 0
for i in range(len(arr)):
if maxes[i] < cutoff:
row = arr[i]
id1,id2 = row['id1'],row['id2']
for sp, cols in spcols:
ogsize_sp = ogs_all[sp]
if (id1 in ogsize_sp and ogsize_sp[id1]>1) or (id2 in ogsize_sp
and ogsize_sp[id2]>1):
for col in cols: arr[i][col] = 0
cleared += 1
print "%s species-sections of rows cleared" % cleared
return arr
def arrfeats_set_gold(arrfeats, pdgold):
arrfeats = ut.arr_copy(arrfeats)
for row in arrfeats:
if pdgold.contains((row[0],row[1])):
row[2] = 1
else:
row[2] = 0
return arrfeats
def norm_columns(arr):
newarr = ut.arr_copy(arr)
for n in newarr.dtype.names:
newarr[n] = scipy.stats.zscore(np.nan_to_num(newarr[n]))
return newarr
def rank_columns(arr):
newarr = ut.arr_copy(arr)
for n in newarr.dtype.names:
newarr[n] = scipy.stats.rankdata(np.nan_to_num(newarr[n]))
return newarr
def rescale_columns(arr, scale_factors):
newarr = ut.arr_copy(arr)
for i,n in enumerate(newarr.dtype.names):
#newarr[n] = np.nan_to_num(newarr[n]/np.max(np.nan_to_num(newarr[n])))
newarr[n] = newarr[n] * scale_factors[i]
return newarr
