Yuki Naito, Project Assistant Professor

Database Center for Life Science (DBCLS)

Interview and text by Nozomi Sakamoto

Advances in DNA sequencing have accumulated vast quantities of sequence data in public databases. Research and development drawing on genome information is under way across the life sciences and medicine, with researchers in Japan and abroad competing closely — and the crux of it is searching and using that enormous body of data efficiently. That calls for software combining speed with ease of use. Yuki Naito, Project Assistant Professor at the Database Center for Life Science (DBCLS), has long worked on technologies for using large-scale nucleotide sequence data, developing software that anyone at the bench can pick up and use immediately: GGRNA, an integrated gene search engine that lets you search genes the way you search Google; GGGenome, which searches nucleotide sequences quickly and without omission; and CRISPRdirect, guide RNA design software for genome editing.

GGGenome, which searches genome and transcript sequences simply and at speed, is published by DBCLS as a tool that runs in a web browser. Every feature is free to use for both commercial and non-commercial purposes. But a web service accessed by many users at once brings constraints — limits on the number of mismatches, insertions and deletions that can be searched, on the maximum number of hits, and on search execution time — as well as the question of keeping a user's query sequence confidential. To address these, Retrieva, Inc. developed and released a package edition that performs the same searches as the web edition on the user's own computer.

This playfully named search tool has also contributed to the safety assessment of two oligonucleotide therapeutics recently approved in Japan, as part of their approval reviews.

1. Review report for Onpattro intravenous infusion 2 mg/mL (2019)

2. Review report for Viltepso intravenous infusion 250 mg (2020)

We asked Professor Naito how GGGenome came to be developed, and what it may mean for the future of life science research.

Sequence search as fast and convenient as Google

— What made you want to develop a nucleotide sequence search tool like GGGenome?

As a graduate student I studied the mechanisms and applications of RNA interference (RNAi), drawing on both molecular biology experiments and computational analysis. RNAi is a landmark technique that uses short double-stranded RNA molecules called siRNA to switch off any given gene, and it is now widely used as an experimental method for elucidating gene function. There were also hopes it could be applied therapeutically by switching off disease-related genes, and the first siRNA drug — Onpattro (generic name patisiran sodium), a treatment for transthyretin amyloidosis — was approved in Europe and the United States in 2018 and in Japan in 2019. Since then, development of oligonucleotide therapeutics applying RNAi (siRNA) to diseases that had been difficult to treat has accelerated.

Life science research involves working with nucleotide sequences constantly. There was already a well-known piece of software for searching them, BLAST, but it was slow, and I was frustrated that short sequences with several mismatches or indels produced incomplete results. Searching nucleotide sequences is not only about finding exact matches; you also need "fuzzy" search that looks for similar sequences containing mismatches and indels. I wondered whether that kind of search could be done as casually as the Google searches we run every day on the internet. Google returns results instantly, so you feel able to keep searching without frustration. I wanted DNA sequences made of the four letters ATGC to be searchable the same way — quickly, instantly, and without missing similar sequences.

Around 2006, while thinking that this was something not just I but probably every life science researcher wanted, I came across Sedue, a text search product developed by Retrieva, Inc. (then Preferred Infrastructure). It searched large volumes of text such as newspaper articles quickly and exhaustively, and I was struck by how fast the results came back. It occurred to me that applying this to nucleotide sequence search could achieve something remarkable. In the life sciences, newly read sequences are registered in public databases and made public by researchers, so I thought it might be possible to build a system for casually searching every sequence accumulated in the world, the way you would google something.

Sedue was designed to search only for exact string matches, however, so we worked with Retrieva to extend it to the fuzzy search that nucleotide sequences require, allowing mismatches and indels — and GGGenome was born. I had wanted a search tool like this myself, and realizing it meant vast data could be used far more efficiently than before. There was an excitement to it: not just making routine work easier, but perhaps leading to new research and new ideas.

— You place great emphasis on how quickly results appear. Why is that?

If a search takes effort and time, you do not feel like casually trying out an idea that has just occurred to you, so I think speed is essential to new discoveries. That is why the other software and tools I develop also aim to produce results quickly. When results appear immediately, you can try all sorts of searches on a passing thought, and a cycle emerges where you look at what came back and run a new search. When search speed differs by a factor of ten or so, I think the way a tool gets used changes as well. There have been many times when I was playing with various sequences in GGGenome myself and made an unexpected discovery. Even when it is not directly useful, it may lead to an idea for new research.

GGGenome in off-target search

— What kinds of research is it used in?

GGGenome is a general-purpose search tool and is genuinely used across a wide range of research, but let me give an example where it is particularly effective: off-target search for oligonucleotide therapeutics and genome editing. Both are powerful methods that can directly target specific RNA or DNA sequences, but one of their challenges is the "off-target effect," where unrelated genes with sequences similar to the intended target are unintentionally affected — which, for a drug, translates into a risk of side effects. Assessing that risk, and then preventing off-target effects, matters a great deal in this field. Inside a living organism, RNA and DNA molecules are in effect performing "sequence search" through base pairing; if we can predict that with string search on a computer, it helps in predicting and avoiding off-target effects.

When I started working on RNAi, I tried off-target search with BLAST, the standard tool for nucleotide sequence search, and noticed the results were incomplete. BLAST's search algorithm was not suited to short sequences of around 20 bases, such as siRNA. Few tools could search short sequences quickly and accurately, and experimental researchers rarely have time to go looking for software, download it and compile it themselves. That is exactly why I wanted to build something that anyone could use straight away and that would search short sequences properly. Being able to start using a tool by glancing at an example, without reading through a manual, matters a great deal to experimental researchers in particular. That is why GGGenome was developed as a web tool.

On the other hand, for users such as pharmaceutical companies developing oligonucleotide therapeutics, a web tool can be awkward when the query sequence has to be kept confidential, and there was also demand for searching unpublished in-house sequence databases. On top of that, because many people access the web edition simultaneously, we impose limits on the number of mismatches, insertions and deletions that can be searched, the maximum number of hits, and search execution time — and there were requests for more detailed searches beyond those limits. That led to the development of the package edition.

The GGGenome package edition combines the fast string search engine developed by Retrieva with the front-end program I developed and published as open source at DBCLS, packaged onto a portable SSD using the container virtualization technology Docker. It runs on Mac, Windows and Linux, and produces the same results as the web edition on the user's own computer regardless of OS or environment. Because it works without an internet connection, it prevents search content from leaking. The package edition also comes with handy scripts for making searches more efficient, and recent updates have added features not present in the web edition, such as gene-of-interest highlighting, developed by Retrieva.

— Is it also used in other fields?

GGGenome is also used as a "component" that handles nucleotide sequence search inside other software. CRISPRdirect, the guide RNA design software for genome editing, runs GGGenome internally in order to predict off-targets. As a tool supporting genome editing by the CRISPR-Cas9 method, CRISPRdirect is being taken up not only in biological research but also in applications to gene therapy and in plant breeding. Plants were originally outside my field, but after CRISPRdirect was published I was contacted by agricultural researchers asking for genome information on major crops to be added. When I added crop genomes to the GGGenome running inside CRISPRdirect, both CRISPRdirect and GGGenome spread among researchers in agriculture.

Computer science contributing to life science R&D

— What is your vision from here?

At present I am developing applications that sequence search makes possible. Off-target search for oligonucleotide therapeutics and CRISPRdirect are two of them, and I am working in parallel on applications that could contribute to other research areas.

With the spread of next-generation sequencers capable of determining large volumes of sequence, nucleotide sequence information is growing explosively, so technologies for analyzing it efficiently will matter more and more. Within that, GGGenome is being used in the safety assessment of oligonucleotide therapeutics and, through the CRISPRdirect application, in genome editing work as well — which gives me great satisfaction as its developer. I intend to keep strengthening its features and to go on developing software and applications that contribute solidly to research and development.

When developing software, it is also important to build in features that can accommodate uses you had not considered and needs you had not anticipated. The reward is that users then try new ways of working, and that feeds back into our own research and development. So I think there is real value in conducting research and development as openly as possible. The cycle of publishing information and receiving feedback leads to new research and new development. When I present at conferences I sometimes meet people using the tools in ways I would never have imagined, and I would like to collect such examples and put them to use in future development. And beyond features that large numbers of users find convenient, I want to keep contributing to niche but interesting research.

10 May 2021 (interview and text by Nozomi Sakamoto)

© 2021 BioImpact Co., Ltd. All Rights Reserved.