def plot_cts_and_outputs(expt_name,
expt_dir,
cts_by_codon_fname,
outputs_fname,
process_label,
tr_codons_fname,
cts_by_codon_xlims=False,
outputs_xlims=False,
log_plot=False):
"""
process_label: string that denotes parameters of the processing used
for tr_codons_fname, including fp size, # genes, and type/# of psct
"""
#Load cts_by_codon and outputs
cts_by_codon = proc.load_cts_by_codon(cts_by_codon_fname)
outputs = proc.load_outputs(outputs_fname)
#Load tr_codon_set
tr_codon_bounds = proc.load_codon_set_bounds(tr_codons_fname)
tr_codon_set = proc.expand_codon_set(tr_codon_bounds)
#load training counts and outputs
cts = proc.get_y(tr_codon_set, cts_by_codon).astype("int")
outputs = proc.get_y(tr_codon_set, outputs)
#Make plots
xlab = "FP counts"
ylab = "# Codons"
title = "{0} Counts by Codon Frequencies".format(expt_name)
xlims = cts_by_codon_xlims
out_fname = expt_dir +\
"/plots/cts_by_codon_freqs.{0}.pdf".format(process_label)
plot.make_cts_by_codon_freq_plot(cts,
title,
xlab,
ylab,
out_fname,
xlims=xlims)
xlab = "Scaled counts"
ylab = "# Codons"
title = "{0} Scaled Counts Frequencies".format(expt_name)
xlims = outputs_xlims
out_fname = expt_dir + "/plots/output_freqs.{0}.pdf".format(process_label)
plot.make_outputs_freq_plot(outputs,
title,
xlab,
ylab,
out_fname,
xlims=xlims,
num_bins=500)
if log_plot:
xlab = "Log scaled counts"
ylab = "# Codons"
title = "{0} Log Scaled Counts Frequencies".format(expt_name)
out_fname = expt_dir +\
"/plots/log_output_freqs.{0}.pdf".format(process_label)
plot.make_log_outputs_freq_plot(outputs,
title,
xlab,
ylab,
out_fname,
num_bins=500)
def make_bin_outputs(expt_dir, outputs_fname, cutoff):
"""
Makes binary transformed outputs file
output = 1 if output >= cutoff else 0
Args:
expt_dir (str) - name of experiment directory
outputs_fname (str) - name of original outputs file
cutoff (float) - scaled counts cutoff to transform to 1, else 0
Returns:
void, just makes binary transformed outputs file w. scaled cts cutoff
"""
outputs = proc.load_outputs(outputs_fname)
bin_outputs = proc.bin_transform_outputs(outputs, cutoff)
bin_out_fname = expt_dir + "/process/bin_outputs.cutoff_{0}.txt".format(
cutoff)
proc.write_outputs(bin_outputs, bin_out_fname)
def make_log_outputs(expt_dir, outputs_fname, scaled_psct):
"""
Makes log transformed outputs file with a scaled pseudocount
Args:
expt_dir (str) - name of experiment directory
outputs_fname (str) - name of original outputs file
scaled_psct (float) - scaled pseudocount to add before log transform
Returns:
void, just makes log transformed outputs file w. scaled psct.
"""
# load outputs
outputs = proc.load_outputs(outputs_fname)
# add scaled pseudocount, log transform outputs
log_outputs = proc.log_transform_outputs(outputs, scaled_psct)
# write log transformed pseudocounts
log_out_fname = expt_dir +\
"/process/log_outputs.scaled_psct_{0}.txt".format(scaled_psct)
proc.write_outputs(log_outputs, log_out_fname)
nn_dir = nn_parent_dir + "/" + model_name +\
"_rep{0}".format(model_rep)
cod_trunc_5p = 20
cod_trunc_3p = 20
min_tot_cts = 200
min_cod_w_cts = 100
# Load transcriptome dicts
len_dict = proc.get_len_dict(gene_len_fname)
cds_dict = proc.get_cds_dict(gene_seq_fname, len_dict)
struc_dict = proc.get_struc_dict(struc_fname)
# Compute cts_by_codon and outputs
cts_by_codon = proc.load_cts_by_codon(cts_by_codon_fname)
outputs = proc.load_outputs(outputs_fname)
paralog_groups = proc.load_paralog_groups(paralog_groups_fname)
# Get gene list to compute performance metrics
gene_set = cts_by_codon.keys()
# Filter genes that are shorter than truncation regions
gene_set = filter(lambda gene: len(cts_by_codon[gene]) > (cod_trunc_5p +
cod_trunc_3p), gene_set)
# Filter genes that don't have enough counts to meet cutoffs
gene_set = filter(lambda gene: proc.has_enough_cts( \
cts_by_codon[gene][cod_trunc_5p:-cod_trunc_3p], min_tot_cts, \
min_cod_w_cts), gene_set)
# Sort genes by footprint density
genes_by_density = proc.sort_genes_by_density(
gene_set, cts_by_codon, cod_trunc_5p, cod_trunc_3p, descend=True)
# NOTE: I'm not sure if I want to filter out paralogs here.