def capture_ancestry(self,
num_of_pcs=2,
covars_to_regress=None,
save_file=None):
logging.info("Running EPISTRUCTURE...")
logging.info("Removing non-informative sites...")
self.meth_data.include(self.informative_sites)
covars = self.meth_data.get_covariates_subset(covars_to_regress)
if covars is not None:
logging.info("Regressing out covariates...")
self.meth_data.regress_out(covars)
else:
logging.info("Ignoring covariates...")
logging.info("Running PCA...")
pca_out = pca.PCA(self.meth_data.data.transpose(
)) # meth_data should be transposed before passing to pca
if save_file:
output_filename = save_file
pcs = pca_out.P[:, range(num_of_pcs)]
data_to_save = column_stack((self.meth_data.samples_ids, pcs))
fmt = '%-12s' + '\t%-12s' * num_of_pcs
savetxt(output_filename, data_to_save,
fmt=fmt) # saves it as samples X PCs
logging.info("The first %s PCs were saved to %s." %
(num_of_pcs, output_filename))
self.components = pcs
def low_rank_approximation(O, k):
"""
O dimensions are n X m
"""
import time
a = time.time()
pca_out = pca.PCA(O)
b = time.time()
res = dot(pca_out.P[:, 0:k], pca_out.U[:, 0:k].transpose())
c = time.time()
logging.debug("PCA TOOK %s SECONDS AND DOT(MULTI) TOOK %s SECONDS" %
(b - a, c - b))
return res
def __init__(self):
logging.info("Testing Started on PCATester")
pca_res_p = loadtxt(self.PCA_P_RES)
meth_data = methylation_data.MethylationDataLoader(
datafile=self.DATA_FILE)
pca_out = pca.PCA(meth_data.data.transpose())
for i in range(10):
assert tools.correlation(pca_out.P[:, i], pca_res_p[:, i])
logging.info("PASS")
logging.info("Testing Finished on PCATester")
def exclude_maxpcstds(self, pcstds):
"""
pcstds is a list of lists (or tuples) where the first index is the pc_index and the second index is the std_num
exclude samples that have std above std_num times higer or lower than the pc std on every std_index
for example, let pcstds be [(1,3),(5,4)] - that will exclude samples that have std above 3 or below -3 on pc 1 and above 4 or below -4 in pc 5
"""
pca_out = pca.PCA(self.data.transpose())
maxpcstds_samples_indices = set()
for pc_index, std_num in pcstds:
logging.info("Finding samples with standard deviation (STD) higher than %d STDs or lower than %d STDs in principal component number %d..." % (std_num, -1 * std_num, pc_index))
pc = pca_out.P[:,pc_index-1] # user start counting from 1 and python from 0
std_pc = std(pc)
maxpcstds_samples_indices.update(where((pc > std_num * std_pc) | (pc < -1 * std_num * std_pc))[0])
if maxpcstds_samples_indices:
logging.info("Excluding samples according to STDs...")
self.remove_samples_indices(list(maxpcstds_samples_indices))
def run(self, args, meth_data):
# run pca and plot PCs
output_filename = args.out if args.out else self.SCATTER_OUTPUT_FILE
try:
assert args.numpcs + 1 < meth_data.samples_size
logging.info("Running PCA...")
pca_out = pca.PCA(meth_data.data.transpose(
)) # meth_data should be transposed before passing to pca
logging.info("Plotting first %s PCs..." % args.numpcs)
pca_scatter_plot = plot.PCAScatterPlot(pca_out,
plots_number=args.numpcs,
save_file=output_filename)
pca_scatter_plot.draw()
except Exception:
logging.exception("In pca plot parser")
raise
def _refactor(self):
"""
run refactor:
exclude bad probes
remove sites with low std
remove covariates
run feature selection
computing the ReFACTor components
find ranked list of the data features
"""
self._exclude_bad_probes()
# self.meth_data.remove_missing_values_sites() # nan are not supported TODO uncomment when supported
self.meth_data.remove_lowest_std_sites(self.stdth)
# self.meth_data.replace_missing_values_by_mean() # nan are not supported TODO uncomment when supported
# feature selection
ranked_list = self._feature_selection()
logging.info('Computing the ReFACTor components...')
sites = ranked_list[:self.t] # take best t sites indices
pca_out = pca.PCA(self.meth_data.data[sites, :].transpose())
score = pca_out.P
logging.info('Saving a ranked list of the data features to %s...' %
self.ranked_output_filename)
ranked_list_output = [
self.meth_data.cpgnames[index] for index in ranked_list
]
savetxt(self.ranked_output_filename, ranked_list_output, fmt='%s')
logging.info('Saving the ReFACTor components to %s...' %
self.components_output_filename)
components = score[:, :self.num_components]
components_output = column_stack(
(self.meth_data.samples_ids, components))
savetxt(self.components_output_filename, components_output, fmt='%s')
return components, ranked_list