is_first_degree = True
is_landmark = False
elif args.nodes_rand > 0:
is_first_degree = False
is_landmark = False
graph_name = "R-{}".format(args.nodes_rand)
else:
is_first_degree = False
is_landmark = False
graph_name = "all_nodes"
# Read in data
try:
assert args.dataset in ['tcga', 'gtex', 'geo']
if args.dataset == 'tcga':
dataset = TCGADataset()
elif args.dataset == 'gtex':
dataset = GTexDataset()
elif args.dataset == 'geo':
dataset = GEODataset(file_path='/network/data1/genomics/D-GEX/bgedv2.hdf5',
seed=seed, load_full=False, nb_examples=40000)
except Exception:
tb = traceback.format_exc()
print(tb)
print("Please enter a valid argument for the dataset. Valid options are tcga, gtex and geo")
import sys
sys.exit()
# Create list of the genes to perform inference on
# If assessing first-degree neighbours, then train only for those genes
from data.datasets import TCGADataset
from data.gene_graphs import GeneManiaGraph, RegNetGraph, HumanNetV2Graph, \
FunCoupGraph
from data.utils import record_result
from tqdm import tqdm
PATH = "/home/user/gil/Expression project/GCN_EXP_predict_TP53_protein_activity/data/gene_tumor_specific"
seed = 0
cuda = torch.cuda.is_available()
graph_dict = {"regnet": RegNetGraph, "genemania": GeneManiaGraph,
"humannetv2": HumanNetV2Graph, "funcoup": FunCoupGraph}
# Read in data: TCGA
dataset = TCGADataset()
# TCGA BRCA samples only
nb_examples = pd.read_csv(
PATH + "/BRCA_samples.csv", # relative python path to subdirectory
header=0, # first row is header.
index_col=0
)
dataset.df = dataset.df.reindex(nb_examples.T.iloc[0])
dataset.sample_names = nb_examples.T.iloc[0].tolist()
# labels: TCGA_BRCA
labels = pd.read_csv(
PATH + "/BRCA_labels.csv", # relative python path to subdirectory
header=0, # first row is header.
index_col=0
)
labels = labels.T.iloc[0]
} # # Select graph and set variables # if args.graph: # # Check graph arg is valid # assert args.graph in graph_dict.keys() # graph_name = args.graph # gene_graph = graph_dict[graph_name]() # is_first_degree = True # else: # is_first_degree = False # graph_name = "all_nodes" # # adj for GCN # adj = gene_graph.adj() # Read in data: TCGA TCGA_dataset = TCGADataset() # tuning num_layer = 2 channels = 30 embedding = 40 dropout = False batch_size = 10 is_first_degree = True # load graph and create adj matrix. graph = "funcoup" gene_graph = graph_dict[graph]() # adj for GCN adj = gene_graph.adj()
# GCN
# model_name = "GCN"
# graph = "humannetv2"
# # create gene graph
# gene_graph = graph_dict[graph]()
# # adj for GCN
# adj = gene_graph.adj()
# is_first_degree = True
# MLP
model_name = "MLP"
graph = "all nodes"
is_first_degree = False
# Dataset: TCGA
dataset = TCGADataset()
# load TCGA BRCA samples
nb_examples = pd.read_csv(
PATH + "BRCA_samples.csv", # relative python path to subdirectory
header=0,
index_col=0)
dataset.df = dataset.df.loc[nb_examples.T.iloc[0]]
dataset.sample_names = nb_examples['x'].tolist()
# labels: load TCGA_BRCA
labels = pd.read_csv(
PATH + "BRCA_labels.csv", # relative python path to subdirectory
header=0, # first row is header.
index_col=0)
labels = labels.T.iloc[0]
dataset.labels = labels.values