Algorithm ramps up genetic computation

'Sailfish' boosts RNA gene expression predictions

DNA

The world has built DNA genomes for a long time, but applying what we know about genetics to everyday medicine is a tough ask.

For example, readers might remember that the business of crafting treatments from genes is so complex that IBM recently entered a partnership to get its Watson megabrain learning to help medicos craft personalised treatments for cancer.

Part of the problem that researchers want to solve is “gene expression”: in all the complexities of how genes interact, what interactions are “expressed” in a physical trait? – whether that trait is blue eyes, or why one individual dies of a cancer that's arrested in someone else.

What's wanted is a way to predict gene expression, and one angle of the research is based on RNA sequencing (RNA-seq) data. The problem is that analysing RNA sequencing is a slow business, and that's where the research out of Carnegie-Mellon University and the University of Maryland comes in. Their Sailfish algorithm dramatically accelerates estimates of the likely outputs of RNA sequence.

To explain why this is important, the researchers' release says: “Though an organism's genetic makeup is static, the activity of individual genes varies greatly over time, making gene expression an important factor in understanding how organisms work and what occurs during disease processes. Gene activity can't be measured directly, but can be inferred by monitoring RNA, the molecules that carry information from the genes for producing proteins and other cellular activities.”

However, analysing the RNA-seq “reads” – short sequences of RNA – traditionally results in huge datasets that have to be mapped back to their original genetic processes. The Sailfish “secret sauce” (except that it's not so secret – the code has been released here) is that it skips this painstaking mapping step.

Instead, the researchers “found they could allocate parts of the reads to different types of RNA molecules, much as if each read acted as several votes for one molecule or another”. Think of it as upvoting posts in a forum: individual votes bestow a kind of consensus on which reads – or posts – carry the greatest significance.

Getting what might be a 15-hour analysis down to minutes is important, the researchers believe: there are already huge repositories of RNA-seq data, but turning data into insight is held back by computational effort.

Fifteen hours for each analysis “really starts to add up, particularly if you want to look at 100 experiments”, explains Carnegie-Mellon associate professor Carl Kingsford. “With Sailfish, we can give researchers everything they got from previous methods, but faster.” ®

Sponsored: Designing and building an open ITOA architecture