NumPy implementation of estimateSizeFactorsForMatrix from DESeq

estimate_size_factors.py

def estimate_size_factor(counts): 
 
    #COUNTS = GENES (rows) x CELLS (columns) 
    genes = counts[:, 0]
    #ONLY COUNTS WITHOUT HEADER NOR GENES
    counts = numpy.array(counts[1:, 1:], dtype = int) 
 
    #DONT REMOVE INFITE VALUES AS IT CAUSES PROBLEMS
    log_counts = numpy.log(counts)
 
    def column_normalization(column):
        distance_to_mean = numpy.log(column) - logmeans
        distance_to_mean = distance_to_mean[numpy.isfinite(distance_to_mean)]
        return numpy.exp(numpy.median(distance_to_mean))
 
    size_factors = numpy.apply_along_axis(column_normalization, 0, counts)
    return size_factors