feat: update spectra-fitting package (#33)

* feat: update spectra-fitting package

BREAKING CHANGES:
kind is moved to shape:{kind}
the options for the optimization package change to optimization object where you can select the kind of method to optimize

* feat: improve coverage

Author: jobo322

package.json

@@ -72,7 +72,7 @@
   },
   "dependencies": {
     "ml-savitzky-golay-generalized": "2.0.2",
-    "ml-spectra-fitting": "^0.6.0",
-    "ml-peak-shape-generator": "^0.9.0"
+    "ml-spectra-fitting": "^0.7.0",
+    "ml-peak-shape-generator": "^0.10.0"
   }
 }

src/__tests__/massPeakPicking.js

@@ -36,7 +36,7 @@ describe('Check the peak picking of a simulated mass spectrum', () => {
       },
     );
     result = optimizePeaks({ x, y }, result, {
-      factorWidth: 1,
+      factorWidth: 4,
       functionName: 'gaussian',
     });
     expect(result[0].x).toBeCloseTo(69.938, 1);

src/__tests__/simulated.js

@@ -11,14 +11,16 @@ describe('Global spectra deconvolution with simulated spectra', () => {
     ];
 
     const data = generateSpectrum(peaks, { from: -1, to: 1, nbPoints: 101 });
-
+    data.y.reverse();
+    data.x.reverse();
     let peakList = gsd(data, {
       minMaxRatio: 0,
       realTopDetection: false,
       smoothY: false,
+      heightFactor: 1,
     });
 
-    let optPeaks = optimizePeaks(data, peakList, {});
+    let optPeaks = optimizePeaks(data, peakList);
 
     expect(optPeaks[0].x).toBeCloseTo(-0.1, 2);
     expect(optPeaks[0].y).toBeCloseTo(0.2, 2);
@@ -27,4 +29,22 @@ describe('Global spectra deconvolution with simulated spectra', () => {
     expect(optPeaks[1].y).toBeCloseTo(0.2, 2);
     expect(optPeaks[1].width).toBeCloseTo(0.1, 2);
   });
+  it('Overlaped peaks', () => {
+    const peaks = [
+      { x: 0.1, y: 0.4, width: 0.0 },
+      { x: 0.101, y: 0.5, width: 0.01 },
+      { x: 0.15, y: 0.4, width: 0.01 },
+      { x: 0.151, y: 0.3, width: 0.03 },
+    ];
+
+    const data = generateSpectrum(peaks, { from: 0, to: 1, nbPoints: 101 });
+
+    let optPeaks = optimizePeaks(data, peaks, {
+      factorWidth: 1,
+      optimization: { kind: 'lm', options: { maxIterations: 300 } },
+    });
+    expect(optPeaks[0].x).toBeCloseTo(0.15, 2);
+    expect(optPeaks[0].y).toBeCloseTo(0.4, 2);
+    expect(optPeaks[0].width).toBeCloseTo(0.01, 2);
+  });
 });

src/gsd.js

@@ -7,6 +7,7 @@ import SG from 'ml-savitzky-golay-generalized';
  * @param {Array<number>} [data.x] - Independent variable
  * @param {Array<number>} [data.y] - Dependent variable
  * @param {object} [options={}] - Options object
+ * @param {object} [options.shape={}] - Object that specified the kind of shape to calculate the FWHM instead of width between inflection points. see https://mljs.github.io/peak-shape-generator/#inflectionpointswidthtofwhm
  * @param {object} [options.sgOptions] - Options object for Savitzky-Golay filter. See https://github.com/mljs/savitzky-golay-generalized#options
  * @param {number} [options.sgOptions.windowSize = 9] - points to use in the approximations
  * @param {number} [options.sgOptions.polynomial = 3] - degree of the polynomial to use in the approximations

src/post/joinBroadPeaks.js

@@ -3,11 +3,20 @@ import { optimize } from 'ml-spectra-fitting';
 /**
  * This function try to join the peaks that seems to belong to a broad signal in a single broad peak.
  * @param {Array} peakList - A list of initial parameters to be optimized. e.g. coming from a peak picking [{x, y, width}].
- * @param {object} [options = {}] -
- * @param {number} options.width - width limit to join peaks.
+ * @param {object} [options = {}] - options
+ * @param {number} [options.width=0.25] - width limit to join peaks.
+ * @param {object} [options.shape={}] - it's specify the kind of shape used to fitting.
+ * @param {string} [options.shape.kind = 'gaussian'] - kind of shape; lorentzian, gaussian and pseudovoigt are supported.
+ * @param {object} [options.optimization = {}] - it's specify the kind and options of the algorithm use to optimize parameters.
+ * @param {string} [options.optimization.kind = 'lm'] - kind of algorithm. By default it's levenberg-marquardt.
+ * @param {object} [options.optimization.options = {}] - options for the specific kind of algorithm.
  */
 export function joinBroadPeaks(peakList, options = {}) {
-  let { shape = { kind: 'gaussian' }, width, lmOptions = {} } = options;
+  let {
+    width = 0.25,
+    shape = { kind: 'gaussian' },
+    optimization = { kind: 'lm' },
+  } = options;
   let broadLines = [];
   // Optimize the possible broad lines
   let max = 0;
@@ -48,7 +57,7 @@ export function joinBroadPeaks(peakList, options = {}) {
               ),
             },
           ],
-          { shape, lmOptions },
+          { shape, optimization },
         );
         let { peaks: peak } = fitted;
         peak[0].index = Math.floor(

src/post/optimizePeaks.js

@@ -6,10 +6,13 @@ import { optimize } from 'ml-spectra-fitting';
  * @param {object} data - An object containing the x and y data to be fitted.
  * @param {Array} peakList - A list of initial parameters to be optimized. e.g. coming from a peak picking [{x, y, width}].
  * @param {object} [options = {}] -
- * @param {string} [options.kind = 'gaussian'] - kind of shape used to fitting, lorentzian, gaussian and pseudovoigt are supported.
- * @param {number} [options.factorWidth = 4] - times of width to group peaks.
+ * @param {number} [options.factorWidth = 1] - times of width to group peaks.
  * @param {object} [options.joinPeaks = true] - if true the peaks could be grouped if the separation between them are inside of a range of factorWidth * width
- * @param {object} [options.optimizationOptions] - options of ml-levenberg-marquardt optimization package.
+ * @param {object} [options.shape={}] - it's specify the kind of shape used to fitting.
+ * @param {string} [options.shape.kind = 'gaussian'] - kind of shape; lorentzian, gaussian and pseudovoigt are supported.
+ * @param {object} [options.optimization = {}] - it's specify the kind and options of the algorithm use to optimize parameters.
+ * @param {object} [options.optimization.kind = 'lm'] - kind of algorithm. By default it's levenberg-marquardt.
+ * @param {object} [options.optimization.options = {}] - options for the specific kind of algorithm.
  */
 
 export function optimizePeaks(data, peakList, options = {}) {
@@ -19,10 +22,8 @@ export function optimizePeaks(data, peakList, options = {}) {
     shape = {
       kind: 'gaussian',
     },
-    optimizationOptions = {
-      damping: 1.5,
-      maxIterations: 100,
-      errorTolerance: 10e-5,
+    optimization = {
+      kind: 'lm',
     },
   } = options;
 
@@ -47,7 +48,7 @@ export function optimizePeaks(data, peakList, options = {}) {
       if (sampling.x.length > 5) {
         let { peaks: optPeaks } = optimize(sampling, peaks, {
           shape,
-          lmOptions: optimizationOptions,
+          optimization,
         });
         for (let j = 0; j < optPeaks.length; j++) {
           optPeaks[j].index = peaks.index;
@@ -68,7 +69,7 @@ export function optimizePeaks(data, peakList, options = {}) {
       if (sampling.x.length > 5) {
         let fitResult = optimize(sampling, [peaks], {
           shape,
-          lmOptions: optimizationOptions,
+          optimization,
         });
         let { peaks: optPeaks } = fitResult;
         optPeaks[0].index = peaks.index;