def _fill_bpscore_matrix(self, verbose=True):
# get unique genes:
all_genes = set(self.assoc.keys())
annotated_genes = set()
for gene, terms in self.assoc.items():
inter = terms.intersection(self.go_dag.terms)
if len(inter) > 0:
annotated_genes.add(gene)
if verbose:
print("Number of annotated genes: " + str(len(annotated_genes))
+ "/" + str(len(all_genes)))
# precompute only for annotated genes, others get a score of 0.0 anyway
genes = annotated_genes
# number of genes
nGenes = len(genes)
# map terms to numberical range [0, nTerms-1]
i = 0
gene_mapping = dict()
idx_2_gene = list()
for t in genes:
gene_mapping[t] = i
idx_2_gene.append(t)
i = i + 1
if verbose:
print("Pre-calculating BPScore between all " + str(nGenes)
+ " genes...")
# float32's precision is sufficient (this way all gene combinations
# can easily be kept in memory)
bpscore_matrix = numpy.zeros((nGenes, nGenes), dtype=numpy.float32)
# fill matrix in row major
for i in range(0, nGenes):
g1 = idx_2_gene[i]
if verbose:
if i % 10 == 0:
print("filling row " + str(i) + "/" + str(nGenes))
for j in range(i, nGenes):
g2 = idx_2_gene[j]
bpscore = GoFastSimilarity.gene_pairwise_score(self, g1, g2)
bpscore_matrix[i][j] = bpscore
bpscore_matrix[j][i] = bpscore
# fill rows sums
row_sums = numpy.zeros(nGenes, dtype=numpy.float32)
for i in range(0, nGenes):
# set the row sum but ignore self
row_sums[i] = numpy.sum(bpscore_matrix[i]) - bpscore_matrix[i][i]
# return the score matrix but also the gene mapping
return (bpscore_matrix, row_sums, gene_mapping)
def _fill_bpscore_matrix(self, verbose=True):
# get unique genes:
all_genes = set(self.assoc.keys())
annotated_genes = set()
for gene, terms in self.assoc.items():
inter = terms.intersection(self.go_dag.terms)
if len(inter) > 0:
annotated_genes.add(gene)
if verbose:
print("Number of annotated genes: " + str(len(annotated_genes)) +
"/" + str(len(all_genes)))
# precompute only for annotated genes, others get a score of 0.0 anyway
genes = annotated_genes
# number of genes
nGenes = len(genes)
# map terms to numberical range [0, nTerms-1]
i = 0
gene_mapping = dict()
idx_2_gene = list()
for t in genes:
gene_mapping[t] = i
idx_2_gene.append(t)
i = i + 1
if verbose:
print("Pre-calculating BPScore between all " + str(nGenes) +
" genes...")
# float32's precision is sufficient (this way all gene combinations
# can easily be kept in memory)
bpscore_matrix = numpy.zeros((nGenes, nGenes), dtype=numpy.float32)
# fill matrix in row major
for i in range(0, nGenes):
g1 = idx_2_gene[i]
if verbose:
if i % 10 == 0:
print("filling row " + str(i) + "/" + str(nGenes))
for j in range(i, nGenes):
g2 = idx_2_gene[j]
bpscore = GoFastSimilarity.gene_pairwise_score(self, g1, g2)
bpscore_matrix[i][j] = bpscore
bpscore_matrix[j][i] = bpscore
# fill rows sums
row_sums = numpy.zeros(nGenes, dtype=numpy.float32)
for i in range(0, nGenes):
# set the row sum but ignore self
row_sums[i] = numpy.sum(bpscore_matrix[i]) - bpscore_matrix[i][i]
# return the score matrix but also the gene mapping
return (bpscore_matrix, row_sums, gene_mapping)
def __init__(self, obo_file, sim_file, mapping_file, sql_conn, verbose=True):
# initialize the super class
GoFastSimilarity.__init__(self, obo_file, sql_conn, verbose)
# load the data file or create and fill it
self._load_or_create(sim_file, mapping_file)
