def main(
data_df,model_df,
start=0,end=None,err=False,coarse_graining_level=0):
dicttype, modeltype = qc.get_model_type(model_df)
seq_cols = qc.get_cols_from_df(data_df,'seqs')
if not len(seq_cols)==1:
raise SortSeqError('Dataframe has multiple seq cols: %s'%str(seq_cols))
seq_dict,inv_dict = utils.choose_dict(dicttype,modeltype=modeltype)
#set name of sequences column based on type of sequence
type_name_dict = {'dna':'seq','rna':'seq_rna','protein':'seq_pro'}
seq_col_name = type_name_dict[dicttype]
#Cut the sequences based on start and end, and then check if it makes sense
if (start != 0 or end):
data_df.loc[:,seq_col_name] = data_df.loc[:,seq_col_name].str.slice(start,end)
if modeltype=='MAT':
if len(data_df.loc[0,seq_col_name]) != len(model_df.loc[:,'pos']):
raise SortSeqError('model length does not match dataset length')
elif modeltype=='NBR':
if len(data_df.loc[0,seq_col_name]) != len(model_df.loc[:,'pos'])+1:
raise SortSeqError('model length does not match dataset length')
col_headers = utils.get_column_headers(data_df)
if 'ct' not in data_df.columns:
data_df['ct'] = data_df[col_headers].sum(axis=1)
data_df = data_df[data_df.ct != 0]
if not end:
seqL = len(data_df[seq_col_name][0]) - start
else:
seqL = end-start
data_df = data_df[data_df[seq_col_name].apply(len) == (seqL)]
#make a numpy array out of the model data frame
model_df_headers = ['val_' + str(inv_dict[i]) for i in range(len(seq_dict))]
value = np.transpose(np.array(model_df[model_df_headers]))
#now we evaluate the expression of each sequence according to the model.
seq_mat,wtrow = numerics.dataset2mutarray(data_df.copy(),modeltype)
temp_df = data_df.copy()
temp_df['val'] = numerics.eval_modelmatrix_on_mutarray(np.array(model_df[model_df_headers]),seq_mat,wtrow)
temp_sorted = temp_df.sort_values(by='val')
temp_sorted.reset_index(inplace=True,drop=True)
#we must divide by the total number of counts in each bin for the MI calculator
#temp_sorted[col_headers] = temp_sorted[col_headers].div(temp_sorted['ct'],axis=0)
MI = EstimateMutualInfoforMImax.alt4(temp_sorted,coarse_graining_level=coarse_graining_level)
if not err:
Std = np.NaN
else:
data_df_for_sub = data_df.copy()
sub_MI = np.zeros(15)
for i in range(15):
sub_df = data_df_for_sub.sample(int(len(data_df_for_sub.index)/2))
sub_df.reset_index(inplace=True,drop=True)
sub_MI[i],sub_std = main(
sub_df,model_df,err=False)
Std = np.std(sub_MI)/np.sqrt(2)
return MI,Std
def main(dataset_df,model_df,left=None,right=None):
# Validate dataframes
qc.validate_dataset(dataset_df)
qc.validate_model(model_df)
# Detect model type based on columns
seqtype, modeltype = qc.get_model_type(model_df)
seqcol = qc.seqtype_to_seqcolname_dict[seqtype]
# Set start and end based on left or right
if not ((left is None) or (right is None)):
raise SortSeqError('Cannot set both left and right at same time.')
if not (left is None):
start = left
end = start + model_df.shape[0] + (1 if modeltype=='NBR' else 0)
elif not (right is None):
end = right
start = end - model_df.shape[0] - (1 if modeltype=='NBR' else 0)
else:
start = model_df['pos'].values[0]
end = model_df['pos'].values[-1] + (2 if modeltype=='NBR' else 1)
assert start < end
# Validate start and end positions
seq_length = len(dataset_df[seqcol][0])
if start < 0:
raise SortSeqError('Invalid start=%d'%start)
if end > seq_length:
raise SortSeqError('Invalid end=%d for seq_length=%d'%(end,seq_length))
#select target sequence region
out_df = dataset_df.copy()
out_df.loc[:,'seq'] = out_df.loc[:,'seq'].str.slice(start,end)
#Create model object of correct type
if modeltype == 'MAT':
mymodel = Models.LinearModel(model_df)
elif modeltype == 'NBR':
mymodel = Models.NeighborModel(model_df)
else:
raise SortSeqError('Unrecognized model type %s'%modeltype)
# Compute values
out_df['val'] = mymodel.evaluate(out_df)
# Validate dataframe and return
return qc.validate_dataset(out_df,fix=True)
def __init__(self,model_df):
"""
Constructor takes model parameters in the form of a model dataframe
"""
model_df = qc.validate_model(model_df.copy(),fix=True)
seqtype, modeltype = qc.get_model_type(model_df)
if not modeltype=='NBR':
raise SortSeqError('Invalid modeltype: %s'%modeltype)
seq_dict,inv_dict = utils.choose_dict(seqtype,modeltype=modeltype)
self.seqtype = seqtype
self.seq_dict = seq_dict
self.inv_dict = inv_dict
self.df = model_df
self.length = model_df.shape[0]+1
# Extract matrix part of model dataframe
headers = qc.get_cols_from_df(model_df,'vals')
self.matrix = np.transpose(np.array(model_df[headers]))
def wrapper(args):
""" Wrapper for function for scan_model.main()
"""
# Prepare input to main
model_df = io.load_model(args.model)
seqtype, modeltype = qc.get_model_type(model_df)
L = model_df.shape[0]
if modeltype == "NBR":
L += 1
chunksize = args.chunksize
if not chunksize > 0:
raise SortSeqError("chunksize=%d must be positive" % chunksize)
if args.numsites <= 0:
raise SortSeqError("numsites=%d must be positive." % args.numsites)
if args.i and args.seq:
raise SortSeqError("Cannot use flags -i and -s simultaneously.")
# If sequence is provided manually
if args.seq:
pos_offset = 0
contig_str = args.seq
# Add a bit on end if circular
if args.circular:
contig_str += contig_str[: L - 1]
contig_list = [(contig_str, "manual", pos_offset)]
# Otherwise, read sequence from FASTA file
else:
contig_list = []
inloc = io.validate_file_for_reading(args.i) if args.i else sys.stdin
for i, record in enumerate(SeqIO.parse(inloc, "fasta")):
name = record.name if record.name else "contig_%d" % i
# Split contig up into chunk)size bits
full_contig_str = str(record.seq)
# Add a bit on end if circular
if args.circular:
full_contig_str += full_contig_str[: L - 1]
# Define chunks containing chunksize sites
start = 0
end = start + chunksize + L - 1
while end < len(full_contig_str):
contig_str = full_contig_str[start:end]
contig_list.append((contig_str, name, start))
start += chunksize
end = start + chunksize + L - 1
contig_str = full_contig_str[start:]
contig_list.append((contig_str, name, start))
if len(contig_list) == 0:
raise SortSeqError("No input sequences to read.")
# Compute results
outloc = io.validate_file_for_writing(args.out) if args.out else sys.stdout
output_df = main(model_df, contig_list, numsites=args.numsites, verbose=args.verbose)
# Write df to stdout or to outfile
io.write(output_df, outloc, fast=args.fast)
def main(model_df, contig_list, numsites=10, verbose=False):
# Determine type of string from model
qc.validate_model(model_df)
seqtype, modeltype = qc.get_model_type(model_df)
seq_dict, inv_dict = utils.choose_dict(seqtype, modeltype=modeltype)
# Check that all characters are from the correct alphabet
alphabet = qc.seqtype_to_alphabet_dict[seqtype]
search_string = r"[^%s]" % alphabet
for contig_str, contig_name, pos_offset in contig_list:
if re.search(search_string, contig_str):
raise SortSeqError("Invalid character for seqtype %s found in %s." % (seqtype, contig_name))
# Create model object to evaluate on seqs
if modeltype == "MAT":
model_obj = Models.LinearModel(model_df)
elif modeltype == "NBR":
model_obj = Models.NeighborModel(model_df)
# Create list of dataframes, one for each contig
seq_col = qc.seqtype_to_seqcolname_dict[seqtype]
L = model_obj.length
sitelist_df = pd.DataFrame(columns=["val", seq_col, "left", "right", "ori", "contig"])
for contig_str, contig_name, pos_offset in contig_list:
if len(contig_str) < L:
continue
this_df = pd.DataFrame(columns=["val", seq_col, "left", "right", "ori", "contig"])
num_sites = len(contig_str) - L + 1
poss = np.arange(num_sites).astype(int)
this_df["left"] = poss + pos_offset
this_df["right"] = poss + pos_offset + L - 1
# this_df[seq_col] = [contig_str[i:(i+L)] for i in poss]
this_df[seq_col] = fast.seq2sitelist(contig_str, L) # Cython
this_df["ori"] = "+"
this_df["contig"] = contig_name
this_df["val"] = model_obj.evaluate(this_df[seq_col])
sitelist_df = pd.concat([sitelist_df, this_df], ignore_index=True)
# If scanning DNA, scan reverse-complement as well
if seqtype == "dna":
# this_df[seq_col] = [qc.rc(s) for s in this_df[seq_col]]
this_df[seq_col] = fast.seq2sitelist(contig_str, L, rc=True) # Cython
this_df["ori"] = "-"
this_df["val"] = model_obj.evaluate(this_df[seq_col])
sitelist_df = pd.concat([sitelist_df, this_df], ignore_index=True)
# Sort by value and reindex
sitelist_df.sort_values(by="val", ascending=False, inplace=True)
sitelist_df.reset_index(drop=True, inplace=True)
# Crop list at numsites
if sitelist_df.shape[0] > numsites:
sitelist_df.drop(sitelist_df.index[numsites:], inplace=True)
if verbose:
print ".",
sys.stdout.flush()
if verbose:
print ""
sys.stdout.flush()
# If no sites were found, raise error
if sitelist_df.shape[0] == 0:
raise SortSeqError("No full-length sites found within provided contigs.")
sitelist_df = qc.validate_sitelist(sitelist_df, fix=True)
return sitelist_df
def main(data_df,
model_df,
start=0,
end=None,
err=False,
coarse_graining_level=0,
rsquared=False,
return_freg=False):
#determine whether you are working with RNA, DNA, or protein.
#this also should determine modeltype (MAT, NBR, PAIR).
dicttype, modeltype = qc.get_model_type(model_df)
#get column header for the sequence column.
seq_cols = qc.get_cols_from_df(data_df, 'seqs')
if not len(seq_cols) == 1:
raise SortSeqError('Dataframe has multiple seq cols: %s' %
str(seq_cols))
#create dictionary that goes from, for example, nucleotide to number and
#visa versa.
seq_dict, inv_dict = utils.choose_dict(dicttype, modeltype=modeltype)
#set name of sequences column based on type of sequence
type_name_dict = {'dna': 'seq', 'rna': 'seq_rna', 'protein': 'seq_pro'}
seq_col_name = type_name_dict[dicttype]
if not end:
seqL = len(data_df[seq_col_name][0]) - start
else:
seqL = end - start
#throw out wrong length sequences.
#Cut the sequences based on start and end, and then check if it makes sense
if (start != 0 or end):
data_df.loc[:,seq_col_name] = \
data_df.loc[:,seq_col_name].str.slice(start,end)
right_length = data_df.loc[:, seq_col_name].apply(len) == (seqL)
if not right_length.all():
sys.stderr.write('''Not all sequences are the same length!
Throwing out incorrect sequences!''')
data_df = data_df.loc[right_length, :]
data_df = data_df.reset_index(drop=True)
if modeltype == 'MAT':
if seqL != len(model_df.loc[:, 'pos']):
raise SortSeqError(
'model length does not match dataset length')
elif modeltype == 'NBR':
if seqL != len(model_df.loc[:, 'pos']) + 1:
raise SortSeqError(
'model length does not match dataset length')
elif modeltype == 'PAIR':
if int(scipy.misc.comb(seqL, 2)) != len(model_df.loc[:, 'pos']):
raise SortSeqError(
'model length does not match dataset length')
#get column names of the counts columns (excluding total counts 'ct')
col_headers = utils.get_column_headers(data_df)
if 'ct' not in data_df.columns:
data_df['ct'] = data_df[col_headers].sum(axis=1)
#remove empty rows.
data_df = data_df[data_df.ct != 0]
#determine sequence length.
#make a numpy array out of the model data frame
model_df_headers = [
'val_' + str(inv_dict[i]) for i in range(len(seq_dict))
]
value = np.array(model_df[model_df_headers])
#now we evaluate the expression of each sequence according to the model.
#first convert to matrix representation of sequences
seq_mat, wtrow = numerics.dataset2mutarray(data_df.copy(), modeltype)
temp_df = data_df.copy()
#evaluate energy of each sequence
temp_df['val'] = numerics.eval_modelmatrix_on_mutarray(
value, seq_mat, wtrow)
#sort based on value
temp_sorted = temp_df.sort_values(by='val')
temp_sorted.reset_index(inplace=True, drop=True)
#freg is a regularized plot which show how sequences are distributed
#in energy space.
if return_freg:
fig, ax = plt.subplots()
MI, freg = EstimateMutualInfoforMImax.alt4(
temp_sorted,
coarse_graining_level=coarse_graining_level,
return_freg=return_freg)
plt.imshow(freg, interpolation='nearest', aspect='auto')
plt.savefig(return_freg)
else:
MI = EstimateMutualInfoforMImax.alt4(
temp_sorted,
coarse_graining_level=coarse_graining_level,
return_freg=return_freg)
#if we want to calculate error then use bootstrapping.
if not err:
Std = np.NaN
else:
data_df_for_sub = data_df.copy()
sub_MI = np.zeros(15)
for i in range(15):
sub_df = data_df_for_sub.sample(int(
len(data_df_for_sub.index) / 2))
sub_df.reset_index(inplace=True, drop=True)
sub_MI[i], sub_std = main(sub_df, model_df, err=False)
Std = np.std(sub_MI) / np.sqrt(2)
#we can return linfoot corrolation (rsquared) or return MI.
if rsquared:
return (1 - 2**(-2 * MI)), (1 - 2**(-2 * Std))
else:
return MI, Std
