pytcpl: A lite Python version of tcpl: An R package for processing high-throughput chemical screening data

Welcome to the GitHub repository for the pytcpl package.

pytcpl is a streamlined Python package that incorporates the mc4, mc5 and mc6 levels of tcpl, providing concentration-response curve fitting and hitcalling based on tcplfit2. It utilizes the Invitrodb version 4.1 release as its backend database.

Pytcpl is at a development stage. If you use it, please contact Eliza Harris (eliza.harris@sdsc.ethz.ch) for assistance and discission. Note that to use pytcpl you will need to install the invitroDB.

---

(Optional) Use conda environment

conda create --name pytcpl

conda activate pytcpl

conda install pip

Install dependencies

pip install -r requirements.txt

Running pipeline:

Notes:

• Download & Install invitroDBv4.1 on you local machine or server
• Goto config_db.yaml to add/set your own MySQL database connection parameters
• Goto config.yaml to customize pipeline behaviour
• Goto DDL to set Data Definition Language (DDL) statements, used to create new MySQL database schema objects
• Automate list generation aeid_list.in suited for balanced workloads for distributed compute instances
• Goto aeid_list.in to set assay endpoint ids (aeids) to be processed by the pipeline (only has effect if aeid_list_manual: 1 in config_db.yaml)

Run pipeline (on single machine)

python src/pipeline/pipeline_main.py --instance_id 0 --instances_total 1

• Replace python with python3 or py depending on your python version
• Goto logs to see the redirected terminal logs and check in the error logs what went wrong for which assay endpoint id
• Note: Parameters instance_id and instances_total are used for distributing workload onto multiple compute engine instances (e.g. gcloud VMs).
• Example: Distribute workload onto 2 VM instances. Run:
  • python src/pipeline/pipeline_main.py --instance_id 0 --instances_total 2 on one machine and
  • python src/pipeline/pipeline_main.py --instance_id 1 --instances_total 2 on another machine

ICE annotation and cytotoxicity estimation

• Generate per-chemical-results from all processed assay endpoints
  • python src/pipeline/pipeline_wrapup.py --instance_id 0 --instances_total 1

Note: Cytotoxicity estimates are made for all assays, but only make sense for non-cell-based target assays. Use hitcall or hitcall_c product (without and with cytotoxicity correction respectively) as needed.

Further automation steps

• Fetch relevant metadata tables from invitrodb
• Remove output files NOT contained in aeid_list.in

Curve surfer: Inspect fitted curves, hitcall labels and potency estimates

streamlit run src/app/curve_surfer_app.py --server.address="localhost"

• If you run the command, the Curve Surfer web app should open as a new browser tab
• Curve Surfer only works for assay endpoints already processed by pipeline
• Or goto online version https://pytcpl.streamlit.app/

streamlit-curve_surfer-2023-08-09-22-08-26.webm