Pages

Monday, April 02, 2012

What happens to compressive sensing when sensing is dirt cheap ?

Two stories blew my mind in the past few weeks. First there was this story told by Elaine Mardis about how she and her team traced back through sequencing that chemotherapy had transformed an already lethal gene of a cancer tumor into a more potent adversary.




 The other story was about  Michael Snyder who using his own blood, figured out through DNA sequencing that a virus had triggered an onset of diabetes 2 much before any of the current testing were capable of detecting it.

It blew my mind for many reasons including how these discovery processes can be accelerated with compressive sensing. Let me clarify why: One of the underlying reasoning as to why compressive sensing is interesting is when the measurement is expensive. Then, fewer measurements means that you can arrive at a solution faster (given the prior knowledge that the solution is sparse).  But in this case, the question we might ask ourselves in the near future is:

What if the cost of DNA sequencing dropped to $1 ?

In this case, the test becomes dirt cheap. The algorithm developed to perform compressed measurements is not that important anymore, as anybody can do the test for a specific disease without the privacy implication of group testing.

But it also changes the whole paradigm. Now, much like  Michael Snyder et al (see below), you have direct access to much data about you or even your tumor. How can compressive sensing help you then since the sensing is supposedly now dirt cheap ? 

You don't realize it yet, because you don't have access to $1 DNA sequencing equipment, but what is now becoming expensive is your time. The time it takes to making sense of changes in your body. If you are checking your DNA every hour, then the sensing of changes, and making sense of them is really the culprit. I simply note from the paper of Snyder et al the following text:

"...We used a Fourier spectral analysis approach that both normalizes the various omics data on equal basis for identifying the common trends and features, and, also accounts for data set variability, uneven sampling, and data gaps, in order to detect real-time changes in any kind of omics activity at the differential time points..."
Aha! uneven sampling, Fourier transform, data gasps, we've seen those before, haven't we ? This entry is listed as both a compressive sensing and a matrix factorization entry.


For those interested, here is the paper:

Personalized medicine is expected to benefit from combining genomic information with regular monitoring of physiological states by multiple high-throughput methods. Here, we present an integrative personal omics profile (iPOP), an analysis that combines genomic, transcriptomic, proteomic, metabolomic, and autoantibody profiles from a single individual over a 14 month period. Our iPOP analysis revealed various medical risks, including type 2 diabetes. It also uncovered extensive, dynamic changes in diverse molecular components and biological pathways across healthy and diseased conditions. Extremely high-coverage genomic and transcriptomic data, which provide the basis of our iPOP, revealed extensive heteroallelic changes during healthy and diseased states and an unexpected RNA editing mechanism. This study demonstrates that longitudinal iPOP can be used to interpret healthy and diseased states by connecting genomic information with additional dynamic omics activity.


Liked this entry ? subscribe to Nuit Blanche's feed, there's more where that came from. You can also subscribe to Nuit Blanche by Email, explore the Big Picture in Compressive Sensing or the Matrix Factorization Jungle and join the conversations on compressive sensing, advanced matrix factorization and calibration issues on Linkedin.

No comments:

Post a Comment