Skip to main content

​
Quick Installation

The simplest way to get started with TRIFID is to install it via pip directly from GitHub: 
pip install git+https://github.com/fpozoc/trifid.git
This method is ideal if you want to:
  • Use TRIFID’s preprocessing modules (QSplice, Pfam effects)
  • Load and analyze pre-computed predictions
  • Integrate TRIFID into your analysis pipeline

​
Development Installation

For development work or to reproduce the complete TRIFID methodology from scratch, follow these steps:
1

Install Conda/Mamba

First, ensure you have Miniconda or Anaconda with mamba installed. If not, install Miniconda:
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
bash Miniconda3-latest-Linux-x86_64.sh -b -p $HOME/miniconda3
Install mamba for faster dependency resolution:
conda install -c conda-forge mamba
2

Clone the Repository

Clone the TRIFID repository and navigate to the directory:
git clone [email protected]:fpozoc/trifid.git
cd trifid
3

Create Environment

Create the conda environment from the provided configuration:
mamba env create -f environment.yml
conda activate trifid
4

Install Pre-commit Hooks

Set up development tools and pre-commit hooks:
pre-commit install

​
Verify Installation

Test that everything is working correctly: 
# Run pre-commit checks
pre-commit run --all-files

# Run tests
pytest -v

​
System Requirements

Minimum Requirements:
  • Python 3.7+
  • 8GB RAM (16GB recommended for genome-wide analysis)
  • Linux or macOS (Windows via WSL)

​
Core Dependencies

TRIFID requires the following Python packages: 
biopython
numpy
pandas>=1.0
scikit-learn
joblib
cython
pyyaml
gtfparse
loguru
muscle

​
Install Optional Dependencies

Depending on your use case, you may want additional features:
For generating plots and interpreting model predictions:
pip install -e .[extra]
This includes: matplotlib, altair, shap, eli5, mlxtend, rfpimp

​
Pre-computed Predictions

Instead of running TRIFID from scratch, you can download pre-computed predictions for multiple genome assemblies. This is the recommended approach for most users.

​
Available Genomes

AssemblyDatabaseVersionRelease DateDownload
GRCh38GENCODEv27Aug 2017Link
GRCh38GENCODEv37Feb 2021Link
GRCh38GENCODEv42Apr 2022Link
GRCh38RefSeq110Feb 2020Link
GRCh37GENCODEv19Dec 2013Link
GRCh37RefSeq105Feb 2020Link
SpeciesAssemblyDatabaseVersionDownload
MouseGRCm39GENCODEM31Link
MouseGRCm38GENCODEM25Link
RatmRatBN7.2Ensembl105Link
ZebrafishGRCz11Ensembl104Link
FruitflyBDGP6Ensembl107Link
WormWBcel235Ensembl108Link
SpeciesAssemblyDatabaseVersionDownload
ChickenGRCg7bEnsembl108Link
ChimpanzeePan_tro_3.0Ensembl104Link
PigSscrofa11.1Ensembl108Link
CowARS-UCD1.2Ensembl104Link
MacaqueMmul_10Ensembl105Link

​
Download and Extract

Each prediction package contains:
  • trifid_predictions.tsv.gz - TRIFID scores for all isoforms
  • trifid_db.tsv.gz - Complete feature matrix
  • Feature description files
# Example: Download human GRCh38 GENCODE 27 predictions
wget https://drive.google.com/... -O trifid_predictions.tsv.gz

# Extract and view
gunzip trifid_predictions.tsv.gz
head trifid_predictions.tsv
Prediction files can be large (>1GB compressed). Ensure you have sufficient disk space.

​
Additional Resources

For advanced usage and model training, you may need additional data files:

Training Data

TRIFID training set (GENCODE 27)

Pre-trained Model

TRIFID model v1.0.4 (pickle format)

Tutorial Notebook

Complete tutorial with examples

Figures Notebook

Reproduce paper figures

​
Need Help?

If you encounter issues during installation:
  1. Check the GitHub Issues for known problems
  2. Ensure all system requirements are met
  3. Try creating a fresh conda environment
  4. Open a new issue with your error log
For species or genome versions not listed, you can request custom predictions by opening an issue on GitHub.

​
Next Steps

Quick Start Guide

Learn how to load predictions and analyze your first gene

Build docs developers (and LLMs) love

Get started for freeTalk to us