UTR_PairPred

Installation

• Install required python libraries with poetry install

Basic requirements

python>=3.9.0
CUDA=11.8
torch=2.2.0

• If you want to preprocess by yourself, please also install tools as following instructions.
  • cd-hit: https://github.com/weizhongli/cdhit
  • ViennaRNA: https://github.com/ViennaRNA/ViennaRNA

Data preprocess

Processed sequence embedding & sequence csv files can be downloaded from here

1. Download GENCODE, Protein-coding transcript sequences fasta file from here

2. Createing sequence df from GENCODE raw fasta file.

cd scripts
sh create_seq_df.sh

• Then, gencode_v44(vM33)_utr_gene_unique.csv and gencode_v44(vM33)_utr_gene_unique_5utr(3utr).fa file will generate.
• If you want to remove similar sequences, please run scripts/cd_hit.sh with those fasta files.

3. Getting sequence embeddings (model inputs).

• With RNA-FM: sh get_emb_rnafm.sh
• With RiNALMo: sh get_emb_rinalmo.sh
• For random forest feature: sh get_rf_feature.sh

Training prediction models

• Use src/run_train_XX.py code for training (replace XX from the below learning method abb table as you want).
• Config also has name rule config/<SPECIES>_<LEARNING_METHOD>.yaml

abb	full
cl	contrastive learning
sv	supervised learning
rf	random forest

• Run example

poetry run python run_train_cl.py --cfg ../config/human_cl.yaml

Downstream analysis

• crossval_analysis.ipynb:
  Performs cross-validation analysis to evaluate the consistency of results across experiments. Visualizes the distribution of cosine similarity and correlations between different experiments.

• sequential_analysis.ipynb: Analyzes basic sequence features (e.g., lengths of 5'UTR, 3'UTR, CDS, and MFE)

• expression_analysis.ipynb: Analyzes translation efficiency (TE) using RNA-seq and Ribo-seq data for each cell line.

Citation

citation information will be written in here