dric2018/EnzymeClassification

Here I am dealing with a bioinformatics problem known as enzyme family classification for a challenge organized by Instadeep and Zindi africa # Enzyme Classification Here I am dealing with a bioinformatics problem known as enzyme family classification for a challenge organized by Instadeep and Zindi africa. Given an amino acid sequence, you have to predict which of 20 classes it is from. As you can see it is an `Sequence classification problem`. # Usage ## Hardware specifications ### Gaming laptop or any computer (e.g workstation) with at least: - 1 NVIDIA GPU (VRAM >= 6 GB since I used a GTX 1060 Max-Q) - 4 Cores CPU (>= Intent core i5 ) - 16 GB of RAM - 50 GB of free space ## Training 1 - Tensorflow version * Create `FOLD` column in `Train.csv` file for cross-validation purpose ``` $ python utils.py --n_fols 5 --data_path ``` * Run `train.py` file with your arguments according to your experiements and your hardware specifications ``` $ python train.py --num_epochs 5 --lr 1e-3 --train_bs 256 --validation_bs 128 --n_folds 5 ``` 2- Pytorch version ## Inference 1 - Load trained weights from here 2 - Run inference script ``` $ python inference.py --batch_size 256 ``` # Disclaimer Some of the utilities used in this repository are based on the ZINDI UMOJAHACK TUNISIA 2020 starter notebook. They have been used as a code base for the experiments and gave a very stable understanding on data preprocessing for this task.