Dateien nach "ILP" hochladen

ILP/nmrSimilarity.py

@@ -1,7 +1,5 @@
-#Also ILP? Minimize distance becuse two vectors might be shifted to oneanother (high weight on height distance to keep same height)
-#Zu hohe Komplexität bei Skalierung
-import gurobipy as gp
-from gurobipy import GRB, Model, quicksum
+import math
+import numpy as np
 
 #Xanthine
 HXANTHINE= {
@@ -51,13 +49,13 @@ C3XANTHINE= {
 }
 
 #7-Methylxanthine
-H3XANTHINE= {
+H7XANTHINE= {
     1: ([7.55], [1]),
     2: ([4.47], [3]),
     3: ([7.72], [1]),
     4: ([7.655], [1]),
 }
-C3XANTHINE= {
+C7XANTHINE= {
     1: ([159.50], [1]),
     2: ([165.47], [1]),
     3: ([122.15], [1]),
@@ -147,16 +145,60 @@ HNMR1= {
 
 
 #Experimental Theobromine nmr
-HNMR1= {
+HNMR2= {
     1: ([11.10], [1]),
     2: ([3.33], [3]),
     3: ([3.84], [3]),
     4: ([7.97], [1]),
 }
 
-def build_model(name, nmrnode, nmrmeasured, excluded_support=None):
-    model = Model(name)
+CNMR2= {
+    1: ([154.9], [1]),
+    2: ([149.8], [1]),
+    3: ([107.1], [1]),
+    4: ([151.0], [1]),
+    5: ([142.8], [1]),
+    6: ([29.3], [1]),
+    7: ([33.9], [1]),
+}
+
+def overlap(listref, listnew):
+    twoleft = np.sum(np.multiply(np.concatenate((listref, [0, 0])),  np.concatenate(([0, 0], listnew))))
+    oneleft = np.sum(np.multiply(np.concatenate((listref, [0])),  np.concatenate(([0], listnew))))
+    neutral = np.sum(np.multiply(listref,listnew))
+    oneright = np.sum(np.multiply(np.concatenate(([0], listref)),  np.concatenate((listnew, [0]))))
+    tworight = np.sum(np.multiply(np.concatenate(([0, 0], listref)),  np.concatenate((listnew, [0, 0]))))
+    overlap = (oneleft + oneright)* 0.5 + neutral
+    return overlap
+
+def bin_array(spectra, highest_ppm, lowest_ppm, bin_width):
+    binnumber = math.ceil((highest_ppm - lowest_ppm)/bin_width)
+    bin = [0] * binnumber
+    for peak in spectra:
+        (shift, height) = spectra[peak]
+        binindex = math.floor((shift[0] - lowest_ppm) / bin_width)
+        bin[binindex] += height[0]
+    normalizedbin = np.divide(bin, np.sum(bin))
+    return normalizedbin
+
+def similarity_nmr(spectraref, spectranew, bin_width, highest_ppm, lowest_ppm):
+
     #Maximize likelihood or minimize Deviation
     #Values for two spectra and optimize largest for both different?
     #Spectra in Nodes to allow maximize overlapp with both spectra or one spectra.
-    #Vergleiche PCA, Binning (Größere Schwankungen zwischen den Methoden)
+    #5.4.2 Eliminating X–H signals from 1H NMR spectra
+    binref = bin_array(spectraref, highest_ppm, lowest_ppm, bin_width)
+    binnew = bin_array(spectranew, highest_ppm, lowest_ppm, bin_width)
+    crosscorr = overlap(binref, binnew)
+    refselfcorr = overlap(binref, binref)
+    newselfcorr = overlap(binnew, binnew)
+    simidx = crosscorr / math.sqrt(refselfcorr * newselfcorr)
+    return(simidx)
+
+def main():
+    print(similarity_nmr(H1XANTHINE, HNMR1, 0.4, 13, 3))
+    print(similarity_nmr(H3XANTHINE, HNMR1, 0.4, 13, 3))
+    print(similarity_nmr(H7XANTHINE, HNMR1, 0.4, 13, 3))
+
+if __name__ == "__main__":
+    main()