Towards Inferring Nanopore Sequencing Ionic Currents from Nucleotide Chemical Structures

The characteristic ionic currents of nucleotide kmers are commonly used in analyzing nanopore sequencing readouts. We present a graph convolutional network-based deep learning framework for predicting kmer characteristic ionic currents from corresponding chemical structures. We show such a framework can generalize the chemical information of the 5-methyl group from thymine to cytosine by correctly predicting 5-methylcytosine-containing DNA 6mers, thus shedding light on the de novo detection of nucleotide modifications. ### Competing Interest Statement The authors have declared no competing interest. * Kmer : DNA or RNA sequence with length of k. Canonical kmer : kmer sequences purely composed of non-modified nucleotides, including {A, T, G, C} for DNA and {A, U, G, C} for RNA. Characteristic ionic current : ionic currents yielded by a specific kmer are usually modeled by a Gaussian distribution, the mean of which is referred to as the characteristic ionic current. Kmer model : a table recording kmers and their corresponding nanopore sequencing characteristic ionic currents. To avoid confusion, the “deep learning model” will be referred to as “framework” throughout the paper. Framework : in this paper “framework” specifically refers to the deep learning model used to predict the characteristic ionic current from kmer chemical structures. GCN : Graph Convolutional Network. CNN : Convolutional Neural Network. NN : Neural Network. RMSE : Root Mean Square Error. R : Pearson correlation. BA : Balanced accuracy. 5mC : 5-methylcytosine. 6mA : N6-methyladenine. I : Inosine. SMILES : Simplified Molecular Input Line Entry System for annotating chemical structures using character strings. Atom : specifically refers to non-hydrogen atoms throughout the paper.