NCBI Seminar

11:15 on Friday Mar 14 

8th Floor Conference Room

 

Alex Churbanov (Host: John Spouge)

 

Clustering ionic flow blockade toggles with Mixture of HMMs

 

Abstract

 

Ionic current blockade signal processing, for use in nanopore detection, offers a promising new way to analyze single molecule properties, 
with potential implications for DNA sequencing. The alpha-Hemolysin transmembrane channel interacts with a translocating molecule in 
a nontrivial way, frequently evidenced by a complex ionic flow blockade pattern with readily distinguishable modes of toggling. 
Effective processing of such signals requires developing unsupervised machine learning methods capable of learning the various 
blockade modes for classification and knowledge discovery purposes. Here we propose method aimed to improve our stochastic analysis
 capabilities to better understand physics of the nanopore channel interactions with analyte. We tailored our memory-sparse distributed 
implementation of Mixture of Hidden Markov Model (MHMM) to the problem of same molecule toggle clustering and analyte classification. 
By using probabilistic fully connected HMM profiles as mixture components we were able to cluster the various 9 base pair hairpin loop toggles.

 

We established that mixture of 12 different toggle profiles of 4 levels produces optimal results both in speed and Maximum a Posteriori (MAP) 
classification accuracy. MAP classification performance depends on several factors such as number of mixture components, number of levels 
in profiles and duration of a toggle sample. Proposed method has many appealing properties like precise classification of analyte in real time,
 unsupervised learning, ability to incorporate new domain knowledge and flexible easily distributable architecture.