Function quantileNormalize

blacklist.cpp:246–294

}

void quantileNormalize(std::vector<std::vector<double>>& data) {
	int cellCount = data.size();
	int binCount = data[0].size();

	//First calculate rank means
	std::vector<double> rankedMean(binCount,0);
	for(int cellID = 0; cellID < cellCount; cellID++) {
		std::vector<double> x(binCount,0);
		for(int binID = 0; binID < binCount; binID++) {
			x[binID] = data[cellID][binID];
		}

		sort(x.begin(), x.end());

		for(int binID = 0; binID < binCount; binID++) {
			rankedMean[binID] += x[binID];
		}
	}
	for(int binID = 0; binID < binCount; binID++) {
		rankedMean[binID] /= (double)cellCount;
	}

	//calculate half value for ties
	std::vector<double> rankedMeanTie(binCount-1,0);
	for(int binID = 0; binID < (binCount-1); binID++) {
		rankedMeanTie[binID] = ((rankedMean[binID]+rankedMean[binID+1])/2);
	}

	//Iterate through each cell line
	for(int s = 0; s < cellCount; s++) {
		std::vector<double> bins(binCount,0);
		for(int p = 0; p < binCount; p++) {
			bins[p] = data[s][p];
		}
		rankify(bins);

		std::vector<double> binsQuantileNormalized(binCount, 0);
		for(int p = 0; p < binCount; p++) {
			if(std::fmod(bins[p],1) != 0) {
				binsQuantileNormalized[p] = rankedMeanTie[(int)floor(bins[p])-1];
			} else {
				binsQuantileNormalized[p] = rankedMean[(int)(bins[p]-1)];
			}

			data[s][p] = binsQuantileNormalized[p];
		}
	}

}