[7a-A503-1~12] 12.6 Nanobiotechnology

[7a-A503-1] Single-molecule Discrimination of Nucleotides through Machin Learnings

〇Masateru Taniguchi^1,2, Takahito Ohshiro^1,2, Takeshi Yoshida^1,2, Takayuki Takaai^1,2, Takashi Washio^1,2 (1.SANKEN, Osaka Univ., 2.JST-CREST)

Keywords:DNA sequencing, machine learnings, single molecules

Current–time profile measurements of the four base molecules of DNA and a solvent were performed using a mechanically controllable break junction. Electric noises were removed from the current–time profiles of the base molecules by learning the characteristics of the electric noises in the solvent. These characteristics were studied using features comprising the electric currents and the times of the current–time waveforms as the vector components. After the noise eliminations, two of the base molecules were identified using the same features and machine learning classifiers, and the F-measure was used as a precision index. As a result, the combinations of features and machine learning classifiers that can achieve discrimination accuracies of 100% were discovered.