This package implements the machine-learning method described by Gao et al (2016) for microbiome classification using a Bayes classifier based on the Dirichlet-Multinomial distribution. In addition to classification, the package also identifies a subset of microbial taxa that can achieve the maximum classification accuracy.
Xiang Gao, Huaiying Lin, Qunfeng Dong (2017); A Dirichlet-Multinomial Bayes Classifier for Disease Diagnosis with Microbial Compositions, mSphere, Volume: 2, Issue: 6.
12/11/2018 v1.1.0 changes include 1) use relative abaundance as input, and in the function, it will be multiple by 10000 and rounding to count data. 2) Will not sum the rest of the variable to create a new vailable called "rest". 3) use Pi score instead of the wilcoxon p value to rank the importantce of features (Xiao Y, Hsiao TH, Suresh U, et al. A novel significance score for gene selection and ranking. Bioinformatics. 2012;30(6):801-7. )
04/12/2020 v1.1.1 1) skip the training set which are not satisfying multinormial distribution. The cases include those subset of your sample which might have all zero in all the selected subset of features. 2) The ten-fold cross validation function was added back.
Please make sure you have devtools installed in your R and do the following:
library(devtools)
install_github("qunfengdong/DMBC")
There are totally 6 functions included in the package. The most important one is dmbc_predict, which will predict probability for a test set from DMBC model given a training set.
MIT
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.