NeoAgDT - Neoantigen Digital Twin

A digital twin software framework for optimal neoantigen-based treatments.

Workflow of NeoAgDT from sequencing data to vaccine composition. Variant/mutation calling and gene expression steps are not part of the pipeline but are shown for general understanding of data flow. NeoAgDT input data is shown as individual boxes (green) in the large box at the bottom left.

Publication

A. Mösch, F. Grazioli, P. Machart, B. Malone: "NeoAgDT: Optimization of personal neoantigen vaccine composition by digital twin simulation of a cancer cell population", Bioinformatics 2024. https://doi.org/10.1093/bioinformatics/btae205

PIP installation

pip install .[all]

Conda installation

Alternatively, we provide an Anaconda-based installation:

conda env create -f environment.yml  # create environment with all dependencies
conda activate neoagdt
pip install .  # install the neoagdt Python package (dependencies should already be insalled at the previous step)

Python 3.9 is required.

---

Generation of input data

Sample commands for running tools that have been used to create the data from Tran et al.

Gene expression with STAR and cufflinks

STAR --genomeDir /path/to/GenomeDir --readFilesIn 1.fastq 2.fastq --outSAMtype BAM SortedByCoordinate
cufflinks -G /path/to/matching/annotation.gtf --library-type fr-firststrand Aligned.sortedByCoord.out.bam

pvacbind including NetChop and NetMHCstab

pvacbind run -b 9999999 --iedb-install-directory /path/to/iedb/ sample.fasta sample samplehlatypes all .
cd MHC_Class_I
pvacbind net_chop --method 20s --threshold 0.0 sample.all_epitopes.tsv sample.fasta sample.all_epitopes.netchop.tsv
pvacbind netmhc_stab sample.all_epitopes.tsv sample.all_epitopes.netmhcstab.tsv 

To create the data needed as input for the neoantigen digital twin, run notebook in notebooks_general/create_dt_input_files_frpm_pvacbind.ipynb Please check the comments and notes in the notebook before running it. The input can be generated by any other script, too, as long as it follows the general requirements (see configuration file)

---

Sample configurations

Configurations are in etc/. Paths can also be absolute and do not need to be relative to the repository.

Simulate a cell population

simulate-cancer-cells etc/cells-config.yaml --logging-level INFO

Run vaccine optimization

The vaccine design optimization supports multiprocessing leveraging on a local dask cluster. It distributes calls to the MIP optimization.

• --num-procs <NUM_PROCS>: number of parallel processes (and of CPU cores)
• --num-threads-per-proc <NUM_THREADS_PER_PROC>: number of threads to allocate for each process. So the total number of threads for a local cluster will be (args.num_procs * args.num_threads_per_proc)

optimize-vaccine-ilp etc/optimization-config.yaml --num-procs <NUM_PROCS> --num-threads-per-proc <NUM_THREADS_PER_PROC> --logging-level INFO

Plot bar charts

create-bar-chart etc/bar-charts-config.yaml --logging-level INFO

Evaluate vaccine response

evaluate-vaccine-response etc/response-likelihood-config.yaml --logging-level INFO

---

Running everything at once

You can run all above-mentioned steps (simulation, optimization, evaluation and plotting) with the following command:

bash ./run-sim.sh

---

Execute unit tests

pip install .[test]
pytest tests

Execute style tests

pip install .[test]
pylint neoag_dt

Documentation

The documentation project for this project can be built with sphinx. The necessary dependencies are install by pip when installing either the all or docs optional dependencies.

pip install .[docs]
cd docs
make html