SEERS (Selective Enrichment of Episomes with Random Sequences)

A systematic exploration of 3'UTR regulatory elements and their contextual associations.

Change Log

Date	Description
2024-08-14	Resolved the issue preventing the model from loading after upgrading TensorFlow to 2.16. Refactored the Jupyter Notebook.
2024-12-08	Added TALE_SNP_effect.ipynb
2025-04-21	Added kmer_motif.ipynb and N45_dissect.ipynb

Data Processing

All paired-end FASTQ files were merged with NGmerge:

./NGmerge -d -1 1.fq.gz -2 2.fq.gz  -o merged.fq.gz

Nn_raw_count.R - Count N45s from the merged FASTQ files.
Nn_nclog2expression.R - Exclude noise from the N45 count results, and infer their regulatory attributes.
ANN_data_prep.R - Prepare data files for model training.

Model Training and Usage

TALE_training_data.tar.bz2 - Training data.
TALE_train.ipynb - Model training and evaluation.
L5-220528_em5-LSTM64x32x0.5-64x0.5-rep4.hdf5 - Our best "context-aware" model (TALE).
TALE_use.ipynb - All in silico experiments.
kmer_motif.ipynb - Generate the motifs in Fig.2.
N45_dissect.ipynb - Generate the heatmaps in Fig.4B.
TALE_SNP_effect.ipynb - Predict 3'UTR variant effect.

K-mer Analyses

kmer_profiling.R - Perform statistical tests for correlations between different k-mers and the regulatory phenotypes.
L5_2-8mer.tsv - Our 2-8 mer profiling result from SEERS.
kmer_profile_visual.R - Visualize the k-mer analysis result.