HmmUFOtu introduction

HmmUFOtu is an HMM based Ultra-fast OTU assignment tool for baterial 16S and amplicon sequencing research, it has two core algorithms, the CSFM-index powered banded-HMM algorithm, and SEP local phylogenetic-placement based taxonomy assignment algorithm.

The main program 'hmmufotu' takes single or paired-end NGS FASTA/FASTQ reads and generate tab-delimited outputs of the taxonomy assignment results of every read. This program supports native multi-threading and SSE2 and up Vectorization.

HmmUFOtu supports all major DNA substitution models, including

• GTR
• TN93
• HKY85
• F81
• K80
• JC69

HmmUFOtu supports variable mutation rate phylogenetic evaluation using the Discrete Gamma (dΓ) models (Yang 1994).

Citations

Zheng, Q., Bartow-McKenney, C., Meisel, J. S., & Grice, E. A. (2018). HmmUFOtu: An HMM and phylogenetic placement based ultra-fast taxonomic assignment and OTU picking tool for microbiome amplicon sequencing studies. Genome biology, 19(1), 82. PubMed ID: 29950165

Implementation

HmmUFOtu is written in pure C++98, and built with the GNU Autotools (autoconfig/automake), and can be easily installed under Linux, Windows and Mac OS X.

Download

You can download the latest release from GitHub at: https://github.com/Grice-Lab/HmmUFOtu/releases. You can clone or fork and pull the source codes from GitHub at: https://github.com/Grice-Lab/HmmUFOtu.

Dependencies

HmmUFOtu depends on the popular head-only C++ libraries Boost and Eigen3. They are available and often pre-installed on most Linux distributions, and can be easily installed on Windows and Mac OS X. The ZLIB and Boost-IOSTREAMS libraries are optionally dependent for handling GZIP/GZIP2 compressed files, but are not required and can be disabled. The JSONCPP C++ library is optionally dependent for formatting HmmUFOtu's assignment output into standard .jplace output; if "jsoncpp" library is found or specified during the configuring step, an optional program hmmufotu-jplace will be installed.

Installation

1. Configure installation, by running the command

./configure

You may consider providing additional options, such as --prefix, --exec-prefix, --with-zlib, --with-boost, --with-jsoncpp, etc.

2. Compile and link, by running the command

make

Look for errors and try to resolve them yourself first (by Google). Contact us only if you are sure it is a bug in our programs.

3. (Optionally) Test, by running the command

make check

It may take a while depending on your processor's speed.

4. Install

make install

You may need root privilege to do it, such as using sudo.

Output

The main program 'hmmufotu' generates tab-delimited tables (TSV files), and is self explanatory. One other major program 'hmmufotu-sum' generates TSV format OTU tables (Operational Taxonomic Tables), which is compatitable with 3rd party tools such as QIIME.

Pre-built databases

You need to build an HmmUFOtu database before assigning taxonomies to your 16S or other target-loci sequencing reads. You can build your own database using hmmufotu-build, or alternatively download the pre-built databases.

Core programs

Core programs are fundamental tools for taxonomy assignment analysis of 16S and other target-amplicon sequencing data. Core programs include:

• hmmufotu-build build an HmmUFOtu database with indexed multiple-sequence alignment (MSA), trained HMM profile, and pre-evaluated phylogenetic tree from reference MSA and tree files
• hmmufotu perform HMM-alignment, phylogenetic-placement based taxonomy assignment for single or paired-end NGS reads
• hmmufotu-sum summarize and generate phylogeny-based OTUs and consensus/prior based OTU representatives by summarizing over multiple assignment results (samples)
• hmmufotu-inspect inspect an HmmUFOtu database, and optionally export its contents

Model training programs

HmmUFOtu distributions contain pre-trained DNA substitution models and Dirichlet density/mixture models required for buiding HmmUFOtu databases. However, the users can train their own models with customized datasets. Model training programs include:

• hmmufotu-train-dm train an HmmUFOtu prior model using Dirichlet Density/Mixture models with customized data
• hmmufotu-train-hmm train a Banded-HMM model with customized data
• hmmufotu-train-sm train a DNA Substitution Model with customized data

Utility programs

Beside its core functionality, HmmUFOtu can perform many additional analysis using the utility programs. Utility programs include:

• hmmufotu-anneal anneal primer sequences to an HmmUFOtu database and evaluate the primer efficiency
• hmmufotu-sim generate simulated single or paired-end NGS reads, aligned or un-aligned, using a pre-built HmmUFOtu database
• hmmufotu-subset subset (subsample) an OTUTable so every sample contains the same mimimum required reads, and prune the samples and OTUs if necessary
• hmmufotu-norm normalize an OTUTable so every sample contains the same number of reads, you can generate a relative abundance OTUTable using a constant of 1
• hmmufotu-merge merge two or more OTUTables, redundant OTUs and samples will be aggregated, an optional merged OTU-tree can also be generated providing the corresponding database
• hmmufotu-jplace format HmmUFOtu's assignment output into standard .jplace file for compatibility of third party tools