Telos-test

This repository is dedicated for the experimentation of Telos. You should have all the packages required for Telos installed. Additionally, you must have rnaseqtools and GFFCompare installed.

Quickstart

You can reproduce the full training, testing, and plotting workflow with the provided script:

./run-all.sh

Notes:

• The script activates a Conda environment named berth. Adjust if your environment name differs.
• Update any paths in your config as needed. The default config folder used is project_config/.

Manual usage

Train on all training data:

python src/train_all_data.py

Test on evaluation datasets:

python src/test_all.py

For both of this scripts, update the paths as necessary.

Generate plots and summaries (aligns with run-all.sh):

• Stage 1 PR curve

  python src/plotters/generate_stage1_pr_curve.py --config_folder project_config/

• Transcript-level PR curve (optionally add --is_train to use training data)

  python src/plotters/generate_transcript_pr_curve.py --config_folder project_config/
  # python src/plotters/generate_transcript_pr_curve.py --config_folder project_config/ --is_train

• Venn diagram of methods (ground truth)

  python src/plotters/plot_venn.py --config_folder project_config/

• Venn diagram of predictions (per model)

  python src/plotters/plot_venn.py --config_folder project_config/ --is_predictions --model_type xgboost
  python src/plotters/plot_venn.py --config_folder project_config/ --is_predictions --model_type randomforest

• For the Jaccard Similarity bar plot

  python src/plotters/plot_venn_barplot.py

• Aggregate transcript level AUPR results across runs

  python src/plotters/gather_auc_results.py --barplot_two_tools

• Plot Stage 1 AuPR Barplots

  python src/plotters/plot_stage1_aupr_barplot.py --generate_all

---

✍️ Author

Developed by Shao Group .

---

For help or issues, open an issue on GitHub or contact the author.