feat: add Probabilistic quotient normalization method (#30)

* feat: add Probabilistic quotient normalization method

* fix: add ml-array-mean

* fix: add ml-array-mean

src/index.js

@@ -93,3 +93,5 @@ export * from './xyObject/xyObjectToXY';
 export * from './zone/zoneToX';
 
 export * from './zones/zonesNormalize';
+
+export * from './matrix/probabilisticQuotientNormalization';

src/matrix/probabilisticQuotientNormalization.js

@@ -0,0 +1,49 @@
+import mean from 'ml-array-mean';
+import { Matrix } from 'ml-matrix';
+
+/**
+ * Performs a Probabilistic quotient normalization over the dataset to account dilution based in median spectrum.
+ * Dieterle, F., Ross, A., Schlotterbeck, G., & Senn, H. (2006). Probabilistic quotient normalization as robust method to account for dilution of complex biological mixtures. Application in 1H NMR metabonomics. Analytical chemistry, 78(13), 4281-4290.
+ * DOI: 10.1021/ac051632c
+ * @param {Matrix} [matrix] - matrix [rows][cols].
+ * @param {Object} [options={}]
+ * @param {number} [options.max=100] - Normalization integral constant.
+ * @return {Object} { data, averageQuotients }.
+ * data: Normalized dataset.
+ * averageQuotients: The quotients of all variables of interest.
+ */
+export function probabilisticQuotientNormalization(matrix, options = {}) {
+  let { max = 100 } = options;
+  matrix = new Matrix(matrix);
+  for (let i = 0; i < matrix.rows; i++) {
+    let row = matrix.getRowVector(i).div(matrix.getRowVector(i).norm() / max);
+    matrix.setRow(i, row);
+  }
+  let normalizationFactor = matrix.norm() / max;
+  matrix.div(normalizationFactor);
+  let referenceSpectrum = [];
+  for (let i = 0; i < matrix.columns; i++) {
+    let currentVariable = [];
+    for (let j = 0; j < matrix.rows; j++) {
+      currentVariable.push(matrix.get(j, i));
+    }
+    referenceSpectrum.push(mean(currentVariable));
+  }
+  let averageQuotients = [];
+  for (let i = 0; i < matrix.rows; i++) {
+    let quotients = [];
+    for (let j = 0; j < matrix.columns; j++) {
+      let quotient = matrix.get(i, j) / referenceSpectrum[j];
+      quotients.push(quotient);
+    }
+    averageQuotients.push(mean(quotients));
+  }
+
+  for (let i = 0; i < matrix.rows; i++) {
+    matrix.mulRow(i, averageQuotients[i]);
+  }
+  return {
+    data: matrix,
+    averageQuotients: averageQuotients,
+  };
+}