About This Project

All documentation can be found here.

Why?

I decided to write a bioinformatics toolchain because even though I am a Quantitative/Computational Biology major, we don't get to work with that much code in class. The things I've done with code so far have mostly been statistics/data related, and I have a lot more fun coding from the ground up. Additionally, I wanted to get more experience coding in CUDA, and algorithms used in bioinformatics are 1. massively parallel and 2. somewhat familiar to me. Finally, I also wanted to gain more experience setting up things like unit/integration tests and CI/CD pipelines.

What's in it?

The code in this repository is written to parse, format/package, and analyze .fasta, .fastq, and .sam files. The regular C++ code supports all 3 of these files. However, the CUDA code only supports operations on .fastq file data. However, all of the CUDA code is neatly wrapped in __host__ code wrappers: you don't actually have to process/format data to use the CUDA kernels I wrote.

Compiling

• Use CMake to compile any projects.
• An example template can be found in the root directory (./CMakeLists.txt). It compiles all CUDA code with all CUDA tests.
• To build, from the root directory:

mkdir && cd build
cmake ..
make
./[EXEC_NAME]

• For compiling with the pure C++ part of this project, you can also just use a makefile. A template can be found at ./tests/cpu/Makefile. It compiles all C++ code with all CPU tests (GTEST).

Developing Environment

• Docker base image: cuda:12.1.1-devel-ubuntu22.04. On my Mac, I use Ubuntu:LATEST since I don't have an NVIDIA GPU on there.
• Docker packages: GTEST, sra-toolkit, valgrind, gdb, cmake, build-essential, and some other not super important ones (see ./Dockerfile).
• On my desktop, I use nvidia-ctk to run CUDA code inside a Docker container with a RTX 4060.