TreeTerminus

TreeTerminus is a program for grouping transcripts into distinct trees for an RNASeq experiment, taking as input Salmon quantified RNA-Seq files. The leaves of the trees represent individual transcripts and internal nodes represent an aggregation of a transcript set. Across the samples, the uncertainty associated with the abundance estimate of the node decreases on an average, ascending the trees.

The transcript trees can be obtained either for each sample or across all the samples in an RNA-Seq experiment. To obtain the summarized trees over all the samples in the RNA-Seq experiment, we provide two modes - Mean and Consensus. The Mean trees are obtained by using mean of reduction in inferential variance across all the samples. To obtain the Consensus trees, first the trees are obtained for each sample and then after processing, are passed as an input to the consensus tree algorithm. PHYLIP's implementation of the majority rule extended tree consensus algorithm has been used.

Building TreeTerminus

TreeTerminus is implemented in Rust and is built on top of Terminus codebase. TreeTerminus uses the cargo build system and package manager. You will need to have rust (ideally v1.54 or greater) installed. To build TreeTerminus, follow the steps below:

git clone [email protected]:COMBINE-lab/TreeTerminus.git
cd TreeTerminus
cargo build --release

To check if the build happens successfully, from the top-level directory of the repository run:

target/release/treeterminus

The above command should not produce any errors and provide the usage information.

Using TreeTerminus

TreeTerminus provides two sub-commands - group and consensus.

The group step is used to generate transcript trees either for each sample or the Mean trees across samples for an RNASeq experiment. The consensus step outputs a set of Consensus trees for the RNASeq experiment using the majority rule extended consensus tree algorithm. It requires that the group step should be run individually on each sample in the experiment.

Group

To obtain the transcript trees for a single sample in an RNA-Seq experiment, run group from the parent directory of TreeTerminus as:

target/release/treeterminus group -d <salmon_dir> -o <out_dir> --mean_inf false

The option -d denotes the path to salmon quantified sample directory, -o denotes the path where TreeTerminus output will be directed, --mean_inf is a boolean indicating whether Mean tree will be constructed. The final trees will be stored in the file group_nwk.txt, inside a subdirectory in the out_dir directory that belongs to a sample in the RNA-Seq experiment.

To obtain the Mean trees for an an RNA-Seq experiment, run group from the parent directory of TreeTerminus as:

target/release/treeterminus group -d <salmon_dir> -o <out_dir> --mean_inf true

Here to -d argument, provide the directory that contains all the salmon quantified samples of interest in the RNA-Seq experiment, rather than just a single sample. The final trees will be stored in the file cluster_nwk.txt, inside the out_dir directory.

The information about the other arguments that can be provided to group, can be obtained by running:

target/release/treeterminus group -h

Consensus

To obtain consensus trees for the RNA-Seq Experiment, run consensus from the parent directory of TreeTerminus as:

target/release/treeterminus consensus -d <salmon_dir> -o <out_dir> 

The inputs to the -d and -o flags are the same as those defined for the group step that was used to generate the Mean tree.

The information about the other arguments that can be provided to consensus, can be obtained by running:

target/release/treeterminus consensus -h

Note - A limitation of the consensus mode only one instance of it can be run at any given moment, aka do not run this on two or more experiments simultaneously.

Example:

Let us assume the following directory structure of a parent directory:

• TreeTerminus
• SalmonQuant
  • SampleA
  • SampleB
  • SampleC
• TreeTermOut

SalmonQuant contains the salmon quantified files for an RNA-Seq experiment, with the individual sample folders inside it being SampleA, SampleB and SampleC. TreeTermOut is where we will keep the output of TreeTerminus.

To run TreeTerminus, first move to its directory as:

cd TreeTerminus

To run group step on a single sample such as SampleA, run:

target/release/treeterminus group -d ../SalmonQuant/SampleA -o ../TreeTermOut --mean_inf false

To obtain the Mean trees for the experiment, run:

target/release/treeterminus group -d ../SalmonQuant -o ../TreeTermOut --mean_inf true

To obtain the Consensus trees for the experiment, run:

target/release/treeterminus group -d ../SalmonQuant -o ../TreeTermOut

Cite

Authors

Noor Pratap Singh, Michael Love, Rob Patro

Downstream Analysis

We have also created an R Package beaveR that parses the output of TreeTerminus, implements DP algorithms for finding an optimal cut, and provides helper functions to obtain useful statistics for subtrees within the TreeTerminus-derived tree structures.

Reproducing the experiments in the paper

The code to reproduce the experiments and figures in the paper are provided at https://github.com/NPSDC/treeterm-paper.