def test_vep_plugin():
gtf = 'tests/data/test.gtf'
vcf = 'tests/data/test.vcf.gz'
fasta = 'tests/data/hg19.nochr.chr17.fa'
dl = SplicingVCFDataloader(gtf, fasta, vcf)
model = MMSplice()
df_python = predict_all_table(model,
dl,
pathogenicity=True,
splicing_efficiency=True)
df_python_predictionsMax = max_varEff(df_python).set_index('ID')
df_plugin = read_vep(vep_output)
df_plugin_predictionsMax = max_varEff(df_plugin).set_index('ID')
indexes = list(
set(df_plugin_predictionsMax.index)
& set(df_python_predictionsMax.index))
vep_plugin_dlogitPsi = df_plugin_predictionsMax.loc[indexes,
'delta_logit_psi']
python_package = df_python_predictionsMax.loc[indexes, 'delta_logit_psi']
assert pearsonr(vep_plugin_dlogitPsi, python_package)[0] >= 0.95
def test_predict_all_table(vcf_path):
model = MMSplice()
dl = SplicingVCFDataloader(gtf_file, fasta_file, vcf_path)
df = predict_all_table(model,
dl,
pathogenicity=True,
splicing_efficiency=True)
assert len(df['delta_logit_psi']) == len(variants) - 1
def test_predict_all_table_exon_dataloader(vcf_path):
model = MMSplice()
df_exons = pd.read_csv(exon_file)
dl = ExonDataset(exon_file, fasta_file)
df = predict_all_table(model,
dl,
pathogenicity=True,
splicing_efficiency=True)
assert len(df['delta_logit_psi']) == df_exons.shape[0]
def test_vep_plugin():
gtf = 'tests/data/test.gtf'
vcf = 'tests/data/test.vcf.gz'
fasta = 'tests/data/hg19.nochr.chr17.fa'
gtfIntervalTree = 'tests/data/test.pkl' # pickle exon interval Tree
dl = SplicingVCFDataloader(gtfIntervalTree,
fasta,
vcf,
out_file=gtfIntervalTree,
split_seq=False,
overhang=(100, 100))
model = MMSplice(exon_cut_l=0,
exon_cut_r=0,
acceptor_intron_cut=6,
donor_intron_cut=6,
acceptor_intron_len=50,
acceptor_exon_len=3,
donor_exon_len=5,
donor_intron_len=13)
df_python = predict_all_table(model,
dl,
batch_size=1024,
split_seq=False,
assembly=True,
pathogenicity=True,
splicing_efficiency=True)
df_python_predictionsMax = max_varEff(df_python).set_index('ID')
df_plugin = read_vep(vep_output)
df_plugin_predictionsMax = max_varEff(df_plugin).set_index('ID')
indexes = list(
set(df_plugin_predictionsMax.index)
& set(df_python_predictionsMax.index))
vep_plugin_dlogitPsi = df_plugin_predictionsMax.loc[indexes,
'mmsplice_dlogitPsi']
python_package = df_python_predictionsMax.loc[indexes,
'mmsplice_dlogitPsi']
assert pearsonr(vep_plugin_dlogitPsi, python_package)[0] >= 0.99
def test_predict_all_table(vcf_path):
model = MMSplice(exon_cut_l=0,
exon_cut_r=0,
acceptor_intron_cut=6,
donor_intron_cut=6,
acceptor_intron_len=50,
acceptor_exon_len=3,
donor_exon_len=5,
donor_intron_len=13)
dl = SplicingVCFDataloader(gtf_file, fasta_file, vcf_path)
df = predict_all_table(model,
dl,
batch_size=1024,
split_seq=False,
assembly=True,
pathogenicity=True,
splicing_efficiency=True)
assert len(df['mmsplice_dlogitPsi']) == len(variants) - 1
def writeMMSpliceToVcf(vcf_in, vcf_lowAF, vcf_out, gtf, fasta):
lowFrequencyVariants(vcf_in, vcf_lowAF)
# dataloader to load variants from vcf
dl = SplicingVCFDataloader(gtf, fasta, vcf_lowAF, tissue_specific=False)
# Specify model
model = MMSplice()
# Or predict and return as df
predictions = predict_all_table(model, dl, pathogenicity=True, splicing_efficiency=True, progress = True)
#generate hash
dict = {}
for row in predictions.itertuples():
id = row.ID
string = exon_sep[0] + "exon:" + row.exons + "," + "delta_logit_psi:" + str(round(row.delta_logit_psi, 4)) + "," + "pathogenicity:" + str(round(row.pathogenicity, 4)) + exon_sep[1]
if id in dict:
dict[id] = dict[id] + string
else:
dict[id] = string
writeTempVCF(vcf_in, vcf_out, dict)
action="store",
dest="vcf",
help="input gzipped vcf")
parser.add_argument("--fasta",
action="store",
dest="fasta",
help="reference genome fasta file")
parser.add_argument("--gtf", action="store", dest="gtf", help="gtf file")
parser.add_argument("--output",
action="store",
dest="output",
help="output file for MMSplice results")
args = parser.parse_args()
# Specify model
model = MMSplice()
#dl = SplicingVCFDataloader(gtf, fasta, vcf, encode=False, tissue_specific=False)
dl = SplicingVCFDataloader(args.gtf, args.fasta, args.vcf)
# Or predict and return as df
predictions = predict_all_table(model,
dl,
pathogenicity=True,
splicing_efficiency=True)
# Summerize with maximum effect size
predictionsMax = max_varEff(predictions)
writeVCF(args.vcf, args.output, predictionsMax)
