Friday, April 15, 2016

Genomic Data: Binary Cuda K-Mer Counting on the GPU: Results

To review, I'm enthralled with the power of GPU and CUDA programming and needed a use-case to "play with" the technology.  My use case is pretty simple, I wanted to see how fast I could count all K-Mers in the full human reference genome using an inexpensive GPU card ($200) on commodity hardware (< $1000).

The results are in, I was able to count all 265,631 unique 9-Mers across the full human genome with 3,209,286,109 9-Mer sequences at over 15 per second.

In other words, in one second, I could get the full count (and locations) of fifteen different 9-Mers across the full 3.2 billion possibilities in the full human genome!

Thursday, April 14, 2016

Counts representing Items and the impact on Data Compression and Encryption

All things in computers are identified in counts of digits (called bits) which are either on or off.  In this post, we'll discuss how counts (discussed in the previous post) are used for identification.