Skip to main content
Fig. 2 | Human Genomics

Fig. 2

From: Prediction of microbial communities for urban metagenomics using neural network approach

Fig. 2

Our general framework. Starting from the map, we simulate the inference task by splitting the samples into the training set (blue dots) and test set (red dots). We use Metaphlan2 [16] to obtain the microbial distribution profiles from the raw sequencing data. We first extract and integrate features for both training and test data. We also construct the evolutionary (phylogenetic) microbial similarity matrix, using the 16s rRNA of the bacteria as a regularizer. Then, we feed the training data’s features and the similarity matrix into MetaMLAnn, which will perform microbial inference based on the features of test dataset. Our model can also be integrated with other classification models trained with same features as an ensemble model

Back to article page