Virtual Nanopore Day, Australia: Bioinformatics

Speaker Details:

James Ferguson, Kinghorn Centre for Clinical Genomics, Garvan Institute

James Ferguson is a Genomic Systems Analyst at the Garvan Institute with a background in clinical pathology testing, algorithms and software development. Leading computational development within the Kinghorn Centre Sydney’s Genomic Technologies Group, James applies his unique skills to develop new bioinformatic tools, as well as design and support nanopore sequencing infrastructure.

Abstract

Signal analysis and deep learning with nanopore squiggles

The raw current level data underlying nanopore sequencing technology has been a treasure trove of richer information when hunting for modifications, higher accuracy, or troubleshooting molecular oddities. However, the management and access of this data is not very simple or straightforward. SquiggleKit was developed to demonstrate and provide tools to allow users to investigate this data, and how to apply algorithms such as structural segmentation and motif identification, as well as file management and manipulation. Deeplexicon builds on these systems by introducing direct RNA barcoding and demultiplexing utilising signal to image conversion and a deep learning model. It is with these tools, we hope to help users create new methods predicated on signal analysis.

Mile Šikić, Genome Institute of Singapore, A*STAR, Singapore/ University of Zagreb, Faculty of Electrical Engineering and Computing, Croatia

Mile Šikić is a group leader at the Genome Institute of Singapore, ASTAR. He is also a professor in computer science at the University of Zagreb. At the same university, he obtained a Ph.D. in computer science in 2008. At the beginning of his career, he was involved, as a system integrator, consultant, and project manager, in more than 70 projects with industry in the fields of computer and mobile networks. In 2009 he became an Assistant Professor. His research interests include developing new algorithms and AI methods for genome sequence analysis and analysis of dynamics in networks.

Abstract

Raven: a de novo genome assembler for long reads

Raven is a straightforward and easy to use tool for long-read de novo genome assembly. The current version is tailored for long reads with an error rate above a few percent, making it particularly suitable for nanopore reads. The performed analysis shows that Raven is one of two fastest de novo assemblers; it reconstructs the sequenced genome in the least amount of fragments, has better or comparable accuracy, and maintains similar performance for various genomes. Raven takes 500 CPU hours to assemble a 44x human genome dataset in only 259 fragments.

Shian Su, WEHI

Shian is a PhD student doing bioinformatics research in the Ritchie Lab at the Walter and Eliza Hall Institute. His focus is on the exploration of DNA methylation using ONT nanopore sequencing, in particular, the visualisation and statistical identification of differentially methylated regions. He has a graduate background in statistics from the University of Melbourne and previously worked in the analysis of RNA-seq and single cell sequencing data. He has developed and contributed to open source Bioconductor projects including Glimma, scPipe, CellBench and NanoMethViz.

Abstract

Visualising Nanopore Methylation Data using NanoMethViz

Novel developments using nanopore sequencing has shown that it is capable of high-resolution, genome-wide detection of CpG methylation in mammalian samples. As the methods for detecting methylation improves, there will be a need for downstream software to analyse this data. We have developed NanoMethViz, a new R/Bioconductor package providing multiple visualisations of nanopore methylation data at different levels of summarisation. It also provides a range of conversion functions to enable to allow the output of various methylation callers to be used with existing differential methylation callers.

Adam Ewing, University of Queensland

Adam Ewing leads the Translational Bioinformatics Group at the Mater Research Institute - University of Queensland. He completed his PhD at the University of Pennsylvania in 2010 working with Prof Haig H. Kazazian Jr, followed by postdoctoral work with Prof David Haussler at the University of California at Santa Cruz. He moved to Australia in 2014 where he has held a DECRA fellowship from the Australian Research Council and is currently the recipient of an Emerging Leadership fellowship from the Medical Research Future Fund. Adam is a collaborative bioinformatician who develops methods and software for analysing high throughput genome sequence data.

Abstract

Mining the Mobile Methylome

Transposable elements (TEs) are mobile segments of DNA that comprise about half of the human genome and contain a subset of active elements that expand in copy number. As TEs are high copy number and are regulated via DNA methylation, the combination of long reads and direct inference of CpG methylation makes nanopore sequencing an ideal method with which to study them. To this end we have developed TLDR (Transposons from Long DNA Reads), which is a long-read insertion detection method that integrates methylation calls to paint a more comprehensive picture of TE structure and regulation than was previously possible.

Zaka Wing-Sze Yuen, The Australian National University

Zaka Yuen is a second-year PhD student at the Australian National University, having graduated with a bachelor's degree in applied science (Honours) majoring in forensics at the University of Canberra. Zaka is interested in exploring the applicability of the Nanopore portable sequencer to forensic genetic and epigenetic polymorphism analysis for the prediction of age and ancestry. Zaka has been working on a benchmarking of various computational tools for the DNA methylation detection from Nanopore sequencing. Their research aims to develop new computational workflows for the simultaneous detection of DNA methylation and genetic variants.

Abstract

Systematic benchmarking of nanopore methylation tools

Various computational tools for modified base detection from Nanopore sequencing have been developed in recent years, but a lack of a systematic benchmarking poses a challenge for users in assessing the reliability of their predictions. Here we benchmarked Nanopolish, DeepSignal, Tombo, Megalodon, Guppy and DeepMod and evaluated the per-site and per-read performance of these tools using controlled methylation mixtures, Cas9-targeted sequencing and bisulfite sequencing. The tested tools showed a trade-off between false positives and false negatives. We explored different ways to improve detection accuracy, including a novel consensus method, METEORE (https://github.com/comprna/METEORE), that combines the outputs from two or more tools.

Eduardo Eyras, The Australian National University

Eduardo Eyras worked at the Sanger Institute (UK) (2001-2004), where he developed a new computational tool to predict splicing variants for Ensembl and contributed to the landmark papers of the mouse, rat, chicken and cow genomes. He was ICREA Professor at the Pompeu Fabra University (Spain) (2005-2019), where he worked on the role of RNA splicing in cancer, and awarded a Young Investigator grant from the European Network on Alternative Splicing (2007). Since 2019, he is EMBL Australia Group Leader at the Australian National University, where he works on new algorithms to study transcriptome variation across individuals and in disease.

Abstract

Reference-free reconstruction and quantification of transcriptomes from Nanopore long-read sequencing

Nanopore sequencing enables the study of transcriptomes from any sample. However, current analysis methods rely on a reference genome or use of multiple sequencing technologies, thereby precluding studies in species with no genome assembly available, in individuals underrepresented in the reference, or to discover disease-specific transcripts not identifiable in the reference. To address these challenges, we developed RATTLE to perform reference-free reconstruction and quantification of transcripts from Nanopore long reads. Using multiple tests and datasets, RATTLE accurately determines transcript sequences and abundances, is comparable to reference-based methods, and saturates in the number of predicted transcripts with increasing input reads.

Shani Amarasinghe, WEHI

Shani completed her PhD in the year 2018 at the University of Adelaide in Plant Bioinformatics before moving to WEHI for her first postdoctoral research position under the supervision of Assoc. Prof Matt Ritchie. At WEHI she works with long-read data and single-cell data in the field of medical informatics. Shani has been in the field of Bioinformatics for over six years. She thoroughly enjoys the challenges she faces as a Bioinformatics scientist and learning to solve problems in analysing high-throughput data in a collaborative environment. Outside of work, she likes swimming, hiking, and entertaining her family and friends.

Abstract

Long-read-tools.org an interactive catalogue of analysis methods for long-read sequencing data

It is challenging for researchers to keep abreast of the rapid growth in software focused on analysing long-read data. In return, this poses the difficulty of selecting the best set of tools to address a particular long-read based research question. long-read-tools.org* is an open-source database that organises the rapidly expanding collection of long-read data analysis tools and allows its exploration through interactive browsing and filtering. long-read-tools.org* enables researchers to help guide the selection of the most relevant tool for their analysis needs and for developers to identify areas of need and existing solutions to benchmark against.

Grace Hall, The University of Melbourne

Grace Hall is a Masters student who completed her research thesis at the Walter and Eliza Hall Institute. Her research focuses on developing software with clinical application. She is currently working on universal DNA-sequencing based tests for bacterial infection, using ONT Nanopore sequencing technology. Alongside research, she has an interest in the communication of, and education in, new discoveries and bioinformatic techniques.

Abstract

Towards Clinical Genomics - Strain Level Identification of Bacterial Infection with NanoMAP

Early and accurate identification of disease is paramount for good patient prognosis. As bacterial infections are common, a rapid, cost-effective, strain-level DNA-based test of bacterial infection is needed. If administered at small medical clinics, it would reduce pressure on the hospital system and allow community monitoring of disease and antibiotic resistance. We present a technological method using ONT products, coupled with novel software, which achieves this capability.

Hasindu Gamaarachchi, Garvan Institute/ UNSW

Hasindu Gamaarachchi completed his bachelor’s degree with first-class honours for Computer Engineering from the University of Peradeniya, Sri Lanka in 2015. Hasindu is in the final stage of his PhD at Computer Science and Engineering, UNSW Sydney. Currently, Hasindu is working as a Genomics Computing Systems Engineer at Garvan Institute of Medical Research. His work is primarily focused on design, development, optimisation and maintenance of bioinformatics software for real-time third-generation sequencing data analysis and prototyping novel domain-specific computer systems for efficient genomics data analysis.

Abstract

F5c for HPC – enabling complete methylation calling pipeline for a PromethION sample in ~15 hours

Methylation calling pipeline for a PromethION sample typically takes days to weeks on a high-performance computer (HPC) when performed using the popular Nanopolish tool. With f5c, a GPU accelerated version of the Nanopolish, this process now takes only ~15 hours. f5c was initially designed to perform real-time methylation calling on embedded systems. Recently, we extended and optimised f5c for high-performance computing environments. With these optimisations, f5c index (compatible with Nanopolish index) takes only ~ 1 hour, f5c call-methylation takes ~10 hours and f5c meth-freq (methylation frequency counting) takes ~0.5 hours on an HPC with 32 CPU cores and an NVIDIA Tesla V100 GPU for a 100 Gbases PromethION sample. With around 3.5 hours spent on alignment using Minimap2 and sorting using Samtools, the total time for the pipeline is thus around 15 hours. f5c is available as open-source software at https://github.com/hasindu2008/f5c/.

Warren Bach, Strategic Relationship Manager, Oxford Nanopore Technologies 

Warren joined ONT in 2019 to support researchers, provide updates as the technology develops and communicate its capabilities across Australia, New Zealand and Singapore. Over 20 years’ experience in both sales and product management roles across a range of genomics technologies including PacBio, Fluidigm, Bionano Genomics, 10X Genomics, Affymetrix etc. Prior scientific career started in biotechnology working in Australia and Japan focusing on recombinant protein expression and purification methods development, scale-up and GMP production.

Stephen Rudd, Project Manager - Bioinformatics, Oxford Nanopore Technologies

Stephen Rudd joined the Product management team last year having previously been the Strategic Account Manager at Oxford Nanopore Technologies for customers in Germany and Austria. Stephen is a classical geneticist and has a background in genome bioinformatics. He has been project manager for a taxonomically diverse range of genome studies utilising most DNA sequencing and genotyping technologies. He is looking forwards to brainstorming potential solutions to challenging problems and to learning more about different research horizons.