Adaptive sampling and real-time in silico enrichment have come up frequently in the last couple of months. I still haven’t used this feature and want to learn more. Tonight I watched the London Calling 2022 session by Alex Sneddon from the Australian National University entitled “RISER: real-time in silico enrichment of RNA species from nanopore signals.” This work was part of Sneddon’s doctoral work. They spoke about how the “extent of splicing complexity remains unknown.” In addition, RNA sequencing plays an important role in elucidating the role of non-coding RNA. Sneddon noted that specialized protocols are expensive, time-consuming, and risk introducing bias. Sneddon also noted the limitations of sequencing cDNA as an RNA proxy. Nanopore sequencing allows for the opportunity to work with RNA directly. Nanopore sequencing sequences RNA from the 3′, which I had not considered. RNA sequencing takes advantage of polyA, and RISER provides real-time RNA sequencing control. RISER’s model uses “convolutional neural network architecture” to take signals and determine if the RNA is coding or non-coding. Sneddon explained that the signal length of four seconds is optimal to determine accurate classification. The software was trained on datasets with 50/50% coding/non-coding sequences. RISER was then tested on a new dataset with high accuracy. RISER is on GitHub and has a simple syntax and output. It can be used for both enrichment and depletion of coding and non-coding sequences.
