Files in this item

FilesDescriptionFormat

application/pdf

application/pdfECE499-Sp2017-zhou-Huozhi.pdf (3MB)
(no description provided)PDF

Description

Title:Noisy mixture models for nanopore genomics
Author(s):Zhou, Huozhi
Contributor(s):Varshney, Lav R.
Subject(s):Noisy mixture models
nanopore genomics
nanopore DNA sequencing
Abstract:We describe a new scenario based on a combination of using nanoscale semi-conductor materials and statistical algorithms to achieve high SNR current signals for robust DNA sequence base calling. In our setting, altered DNA molecules are threaded through nanopores in electrically active two- dimensional membranes such as graphene and molybdenum di-sulphide to be sensed by changes in electronic currents owing through the membrane. Unfortunately, solid-state nanopores have been unsuccessful in DNA base identification due to the conformational stochastic fluctuations of DNA in the electrolytic solution inside the pore, which introduces signifcant noise to the measured signal. Hence, we propose an integrated effort that combines electronic simulation based on device physics with statistical learning algorithms to perform clustering and inference from the solid-state nanopore data. In particular we develop Gaussian Mixture Models (GMMs) that take into account the characteristics of the system to cluster the electrical current data and estimate the probability of the DNA position inside the nanopore. The validity of the learning algorithms for noisy GMM model has been demonstrated for uniform and Gaussian noise models with synthetic data sets. We also demonstrate the implementation of a pipelined version of the GMM training algorithm, which can be used to realize in near-sensor computing and inference systems. Finally, we also propose one possible solution to the theoretical resolution limit of nanopore DNA sequencing.
Issue Date:2017-05
Genre:Other
Type:Text
Language:English
URI:http://hdl.handle.net/2142/97935
Date Available in IDEALS:2017-09-07


This item appears in the following Collection(s)

Item Statistics