Classification of GC and health control samples based on LDA model.
For Classification, The pre-trained model was saved in the bestLDA.Rds file. The model can be read as below:
lda.mod = readRds('./bestLDA.Rds')
Then, any new metabolism matrix with the 19 metabolites (LPC 17:0,PE O-44:6|PE O-24:2_20:4,CAR 14:0,HexCer 42:2;2O|HexCer 18:1;2O/24:1,FA 28:3;O_Neg,HexCer 42:2;3O_iso2_Neg,LPE O-14:1_Neg,PE 36:4|PE 18:2_18:2_Neg,CAR 18:1,Cer 38:1;2O|Cer 18:1;2O/20:0_Neg,TG(P)50:2,FA 18:0;O_Neg,PC O-38:3,FA 16:2_Neg,PC O-42:6|PC O-22:2_20:4,Hex2Cer 42:2;2O|Hex2Cer 18:1;2O/24:1,PE 40:7|PE 18:1_22:6_Neg,LPE 20:4,PE 40:3_Neg) as the colunms and the sampleID as the columns can be read into R as one matrix: idt.data; the prediction can be done by the following codes:
input.items = colnames(lda.mod$trainingData)[-1]
idt.data = log2(as.matrix(idt.data))+5
idt.data = scale(idt.data)
ind.y = predict(lda.mod,idt.data[,input.items],type = 'prob')
The code was also in predict_LDA.R