cadencorontzos / fastdna

Line-rate DNA direct encoder

FastDNA 4:1 encoder

The fastdna program can encode DNA ASCII files into binary files that are one quarter of the original file, or decode them back. The raw inputs are ASCII files that contain only the bases (characters) 'A', 'C', 'G', or 'T'. The encoded files (.2b extension) are binary files, which each base encoded with a unique two-bit combination.

The program has various options to control its speed. Use ./fastdna -h to list all these options.

Building it

The program is designed for x86-64 architectures, ideally with the BMI2 instruction set. To build it on such an architecture, simply type make. You may want to override the compiler choice using CXX, e.g., make CXX=clang++-12.

Testing it

The script runtests.sh will go through a set of simple combinations of inputs and parameters. If all the tests pass, you won't see any error messages.

Measuring it

Evaluating the full performance potential of fastdna requires large inputs sizes (in the order of GBs.). But because the contents of the input file don't affect performance, you can generate a random input for benchmarking. One easy way to generate a random file (say, of size 1TB), might be:

tr -dc 'ACGT' < /dev/urandom | head -c 1T > inputfile

Rebuilding the document

The document cal.pdf can be built from the source markdown file cal.Rmd. It requires R with a few packages, as well as the cal.csv data file with the results of the experiment. Both cal.csv and cal.pdf can be built from scratch using make once the prerequisites have been met.