Taiyaki and mRNA m5C

Retraining a Nanopore model

Taiyaki is a model-retraining algorithm developed by Oxford Nanopore. This software is very useful in re-train Nanopore models from raw signals to recognize modified bases.

It a very interesting experience for me to seek for a Nanopore model can decipher mRNA m5C directly. In our hands, we used modified and unmodified in vitro transcripts to construct the training set, then process them to Taiyaki. Unfortunately, we found that the model was easily overfit.

Seems good, but no m5C site was found correctly in the real data.



Back