2024ibis-challenge

IBIS Challenge code scripts for Pap Team.

G2A-AAA

Pre-processing notes

HT-Selex: flexbar was used to process .fastq files, for example:

flexbar --reads data/HTS/NACC2/NACC2_R0_C2_lf5ACGACGCTCTTCCGATCTTG_rf3TGTGTTAGATCGGAAGAGCA.fastq --post-trim-length 40 --min-read-length 20 --qtrim-threshold 30 --output-reads HTS_NACC2_R0_C2.fasta --fasta-output --number-tags --stdout-log > log_flexbar2.txt

Calling bed2ntohe.py $input_data converts the file to an OHE nucleotide representation and saves it as a numpy array to disk.

Creating the entities of the negative class and forming the validation set are done by:
prepare_training.py $input_file

Model training

The cnn_01.py contains code for model creation and fitting.

Submission scripts

The prep_preds_subm.py and predict_final.py are used for preparing the submissions. This includes loading the target test input data, and the corresponding tags. Then getting the probabilities via model inference.