Metagenome Prediction Tutorial

Introduction

This tutorial explains how to predict a microbial community metagenome using PICRUSt, based on 16S (or other marker gene) data as detailed in Picking OTUs for use in PICRUSt.

BIOM Format

  • Please note that PICRUSt by default uses the relatively new biom format for representing OTU tables and Gene tables (e.g. KOs by samples). This has several benefits including easier integration with other software (e.g. QIIME and others in the future) and allows embedding of extra metadata about both the samples and observations (OTUs/KOs).
  • However, PICRUSt also allow users to input OTU tables and export PICRUSt predictions in tab-delimited format by using the ‘-f’ option (see below).
  • In addition, users can always convert to/from biom format to tab-delimited format using BIOM’s built in conversion script.

Requirements

  1. You should have already installed PICRUSt (Installing PICRUSt).
  2. A PICRUSt compatible OTU table (Picking OTUs for use in PICRUSt), such as the example file tutorials/hmp_mock_16S.biom. Note that this OTU table was reference picked against Greengenes version 18may2012.

Step 1: Normalize OTU Table

normalize_by_copy_number.py normalizes the OTU table by dividing each OTU by the known/predicted 16S copy number abdundance.

Input is the users OTU table (that has been referenced picked against Greengenes).

Input and output files are in biom format:

normalize_by_copy_number.py
        -i your_otu_table.biom
        -o normalized_otus.biom

(Optional) Previous examples assume the most recent Greengenes was used for closed OTU picking. Older versions can be specified using the --gg_version option:

normalize_by_copy_number.py
      --gg_version 18may2012
      -i hmp_mock_16S.biom
      -o normalized_otus.biom

Step 2: Predict Functions For Metagenome

predict_metagenomes.py creates the final metagenome functional predictions. It multiplies each normalized OTU abundance by each predicted functional trait abundance to produce a table of functions (rows) by samples (columns).

Input is the normalized OTU table created by normalize_by_copy_number.py.

Output is in biom format by default:

predict_metagenomes.py
        -i normalized_otus.biom
        -o metagenome_predictions.biom

(Optional) Output format can be changed from BIOM to tab delimited using ‘-f’ option:

predict_metagenomes.py
        -f
        -i normalized_otus.biom
        -o metagenome_predictions.tab

(Optional) NSTI values for each sample can be obtained using the -a option. (We strongly recommend this step, as NSTI values are precalculated for common inputs):

predict_metagenomes.py
        -i normalized_otus.biom
        -o metagenome_predictions.tab
        -a nsti_per_sample.tab

(Optional) Previous examples assume the most recent GreenGenes was used for closed OTU picking. Older versions can be specified using the --gg_version option:

predict_metagenomes.py
        --gg_version 18may2012
        -i normalized_otus.biom
        -o metagenome_predictions.biom

(Optional) Previous examples assume that KEGG Orthologs predictions are wanted. Other types of functions (e.g. COGs) can be specified using the --type_of_prediction option:

predict_metagenomes.py
      --type_of_prediction cog
      -i normalized_otus.biom
      -o metagenome_predictions.biom

Step 3: Analysing Predicted Metagenomes

Now that you have your predicted metagenome you can analyse it just the same as a real metagenome.

Some suggestions for downstream analysis are outlined in Analyzing PICRUSt predicted metagenomes.