MetabAnnotate

Annotate pathway labels for unknown metabolites

Load example data.

library("MetabAnnotate")
data("exampleData")

Step 1. Estimate the number of latent factors.

We recommend using BEMA for raw data, and dBEMA for summary-level data.

The BEMA function is part of the package and was downloaded from the GitHub repository maintained by ZhengTracyKe, available at: https://github.com/ZhengTracyKe/BEMA.

Y.copy = na.omit(Y)#remove rows including NAs in Y
K = BEMA(data = t(Y.copy),n = ncol(Y.copy),p = nrow(Y.copy),alpha = 0.1)

The dBEMA function is part of the package and was obtained from the GitHub repository maintained by Weiqiong Huang, available at: https://github.com/WeiqiongHuang/HiGSS.

est.K = dBEMA(Stand.B = B,N = N,alpha = 0.1,n.avg = 5)# Only 5 simulations to estimate the bulk eigenvalue distribution for time efficiency.
K = est.K$nFactors

Step 2. Transform the input data into the format required by MetabAnnotate.

Raw data

mtb_data = preProcess(IDs = IDs,Y = Y,K = K,impute = T,imputation.methods = "Mean",COMD.column = "CHEMICAL_ID")

Summary level data

mtb_data = preProcess(IDs = IDs,B = B,N = N,K = K,COMD.column = "CHEMICAL_ID")

Step 3. Implement the annotation function

Implement the annotation function by randomly selecting 20 metabolites for calibration. Generate 20 Gibbs samples, discarding the first 10 iterations as burn-in. It is recommended to run 100 iterations and 50 burn-in for real annotation tasks (Iter = 100, burnin = 50).

test = metabAnnotate(mtb_data = mtb_data,size.clbr = 20,Iter = 20,burnin = 10,new.pathway = F,seed = 1)

Check the annotation results.

head(test$annt.result)