ONT-AutoEncoder

This is a repository with code and Jupyter Notebooks about experiments with deep AutoEncoders on Oxford NanoPore data. Several different deep AutoEncoders models were built for the detection of m6A modification on raw current signals from direct-RNA sequencing experiements using ONT. The models were trained and tested on two datasets: the first one is an in-silico generated dataset made of ONT-like raw currents produced by two simple computational models of the ONT pore. Then, the same models were challenged against the real-world Epinano-Curlcakes dataset produced by Liu et al., in the 2019. The "culcakes" dataset was composed of raw currents generated by MinION ONT platfrom by sequencing synthetic constructs (curlcakes) containg or not, the m6A modified base with well known positions.