New basecaller now performs ‘raw basecalling’, for improved sequencing accuracy

Oxford Nanopore has released a new basecaller, Albacore v2.0.1, that identifies DNA sequences directly from raw data, rather than utilising an intermediary stage called ‘event detection’. This upgrade enhances accuracy of the single-read sequence data, contributing to high consensus accuracy for nanopore sequence data.

Moving to raw basecalling also opens up opportunities for further improvements in sequencing speed (throughput), signal-to-noise and RNN features.

This new basecaller release also enables resolution of longer homopolymers for the 1D^2 sequencing method, a feature that has been available for 1D basecalling since Albacore 1.0 (March 2017) and integrated in MinKNOW 1.6 (May 2017).

The accuracy of nanopore sequencing data has been driven by continuous releases that update sequencing chemistry and data analysis methodology.
Performance evolution nanopore.JPG
For example, moving from Hidden Markov Model (HMM) to Recurrent Neural Nets (RNN) for basecalling provided a substantial improvement in single-read and consensus accuracy, alongside other releases such as the R9 series nanopore.  The release of raw basecalling is the next step in our data analysis techniques; higher fidelity data can be obtained from the original, raw nanopore signal, particularly when operating at faster sequencing speeds.

Raw basecalling is available now for 1D and future updates will make it available for 1D squared and RNA sequencing.

Read the community post. (Customer login required)