Dresden 2020 – wissenschaftliches Programm
Die DPG-Frühjahrstagung in Dresden musste abgesagt werden! Lesen Sie mehr ...
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
BP: Fachverband Biologische Physik
BP 19: Poster VII
BP 19.21: Poster
Dienstag, 17. März 2020, 14:00–16:00, P2/3OG
Deep Learning for DNA Reads through 2D Solid-state Nanopores — •Angel Diaz Carral1, Chandra Shekar Sarap1, Ke Liu2, Aleksandra Radenovic2, Maria Fyta1, and Elka Radoslavova3 — 1Institute for Computational Physics, Universität Stuttgart, Germany — 2Laboratory of Nanoscale Biology, Institute of Bioengineering, School of Engineering, EPFL, Switzerland — 3Fundamental Physics Department, Faculty of Science UNED, Spain
DNA molecules can electrophoretically be driven through nanopores giving rise to measurable electronic current blockades important for DNA sensing. In this work, experimental ionic traces from 2D molybdenum disulfide nanopores DNA translocations are used to train a Deep Learning model for interpreting the DNA events and improving the identification of different nucleotides threading the nanopore. We propose a methodological approach to train a Clustering model for identifying molecular events related to different conformations of DNA nucleotides threading the nanopore. This approach has revealed the efficiency of using the height of the ionic blockade as the training feature. This can lead to a clear clustering of the nanopore events over the use of the traditionally used dwell time. In order to eventually predict the type of molecule threading the pore, the features selected in the clustering analysis are used to train a Convolutional Neural Network capable of optimizing and accelerating the nanopore read-out. Our approach allows for a deep insight into characteristic molecular features in 2D nanopores and provides a feedback mechanism to tune these materials and interpret the experimentally measured signals.