Deep-learning-based approach for identification of eukaryotic sequences in the metagenomic data powered by PyTorch.

The sequences are classified in two stages:

• In the first stage, the sequences are classified to classes: archaea, bacteria, prokarya, eukarya, organelle and unknown.
• In the second stage, the sequences labeled as organelle in the first stage are classified to either mitochondria, plastid or unknown.

For more information, please refer to our paper: Tiara: Deep learning-based classification system for eukaryotic sequences.

Supplementary data

Supplementary sequences

• Python >= 3.7, <=3.9
• numpy, biopython, torch, skorch, tqdm, joblib, numba

More detailed installation instructions can be found here.

Run pip install tiara, preferably in a fresh environment.

Run conda install -c conda-forge tiara, preferably in a fresh environment.

We recommend to use mamba instead of conda (it's faster).

Unfortunately currently it does work only for python 3.7 and 3.8.

• Download latest release from https://github.com/ibe-uw/tiara/releases.
• Unzip/untar the archive.
• Go to the directory.
• Run python setup.py install.

git clone https://github.com/ibe-uw/tiara.git
cd tiara
python setup.py install

After the installation, run tiara-test to see if the installation was successful.

tiara -i sample_input.fasta -o out.txt

The sequences in the fasta file should be at least 3000 bases long (default value). We do not recommend classify sequences that are shorter than 1000 base pairs.

It creates two files:

• out.txt, a tab-separated file with header sequence id, first stage classification result, second stage classification result.
• log_out.txt, containing model parameters and classification summary.

tiara -i sample_input.fasta -o out.txt --tf mit pla pro -t 4 -p 0.65 0.60 --probabilities

In addition to creating the files above, it creates, in the folder where tiara is run, three files containing sequences from sample_input.fasta classified as mitochondria, plastid and prokarya (--tf mit pla pro option).

The number of threads is set to 4 (-t 4) and probability cutoffs in the first and second stage of classification are set to 0.65 and 0.6, respectively.

The probabilities of belonging to individual classes are also written to out.txt, thanks to --probabilities option.

For more usage examples, go here.

Michał Karlicki, Stanisław Antonowicz, Anna Karnkowska, Tiara: deep learning-based classification system for eukaryotic sequences, Bioinformatics, Volume 38, Issue 2, 15 January 2022, Pages 344–350, https://doi.org/10.1093/bioinformatics/btab672

Tiara is released under an open-source MIT license

• 1.0.3 – added pyproject.toml, updated dependencies to python<3.10 – unfortunately tiara doesn't work right now with python newer than 3.9 due to torch 1.7.0 compatibility issues. Added option to use gzipped fasta file as input (automatically identified by .gz suffix).
• 1.0.2 – added Python 3.9 compatibility, added an option to gzip the results. Added this README section.
• 1.0.0, 1.0.1 – initial releases.