---
title: 'Application of genome graph for standard representation of structural variations in RDF'
title_short: 'Structural variations in RDF'
tags:
- structural variations, genome graph, knowledge graph
authors:
- name: Toshiaki Katayama
orcid: 0000-0003-2391-0384
affiliation: 1
- name: Takatomo Fujisawa
affiliation: 2
- name: Yuki Moriya
affiliation: 1
- name: Shuichi Kawashima
affiliation: 1
- name: Mayumi Kamada
affiliation: 3
affiliations:
- name: Database Center for Life Science, Research Organization of Information and Systems, Kashiwa, Chiba, Japan
index: 1
- name: National Institute of Genetics, Mishima, Shizuoka, Japan
index: 2
- name: Department of Biomedical Data Intelligence, Graduate School of Medicine, Kyoto University, Kyoto, Japan
index: 3
date: 12 November 2021
bibliography: paper.bib
event: Spain2021
biohackathon_name: "ELIXIR BioHackathon 2021"
biohackathon_url: "https://biohackathon-europe.org/"
biohackathon_location: "Barcelona, Spain, 2021 (remote)"
group: Application of genome graph for standard representation of structural variations in RDF
# URL to project git repo --- should contain paper.md
git_url: https://github.com/elixir-europe/bioHackathon-projects-2021/tree/main/projects/31
# This is the short authors description that is used at the
# bottom of the generated paper.
authors_short: Toshiaki Katayama & Takatomo Fujisawa \emph{et al.}
---

# ELIXIR BH21 SV-RDF

Authors
* Toshiaki Katayama
* Takatomo Fujisawa
* Yuki Moriya
* Shuichi Kawashima
* Mayumi Kamada

## Introduction

Representation of structural variations (SVs) is one of the urgent issues for data sharing. Some databases already use the VCF format for this purpose. However, the way to serialize SVs is not standardized and the format will not be applicable for more complex SVs detected in the future. Meanwhile, the 'vg' tool that constructs and manipulates genome graphs has the capability to serialize any variations into the Resource Description Framework (RDF). In this proposal, we plan to apply the generic 'vg' RDF model for representing existing SVs to be standardized and extended for future complex SVs. As we have been developing knowledge graphs in RDF covering a diverse range of life sciences and biomedical datasets, the interpretation of SVs can be seamlessly realized.

## Methods

https://docs.google.com/document/d/1tG9Zv0wzwvZVYGqiOaqsmli1u3U8x14NSboUApEwVyg/edit#heading=h.r3tromt0fbkg

課題整理

* 保存されている両端のノード
* ref/alt の入れ替えた場合の対称性
* rdf:value に配列が必要？
* FALDO Region はなぜ向きを持たないのか
* 代わりに begin, end それぞれに Fwd/Rev Position を持たせている
* なぜか。富山BHを思い出しつつ、たぶん position に fwd か rev かそれ以外を書く必要のあるユースケースがあって、region に向きをもたせる場合と begin/end のポジションにもたせる場合で SPARQL が変わるのはイヤとか？
* いずれにしても begin/end の向きをみれば region の向きは推論できる
* SV の階層性
* caller ごとに複数の結果が出ていて、それをマージして variant region をサブミッションにしている場合あり

### VCF to (GFA to) RDF

* https://github.com/dbcls/visc/wiki/VISC-SPARQLthon107
* https://github.com/dbcls/visc/wiki/VISC-SPARQLthon109

## Results

SVはサブミッターのアノテーション、主張（ファクトではない）

目的：サブミッションをうけつけたい、ゲノムグラフでそれができればキレイ、無理なら？

現実:各SV callerが作った機械的な結果も、マニュアルキュレーションされた結果も、両方格納する必要あり（整合性が取れなかったとしても）

なので、オントロジーファースト、ゲノムグラフファーストはリアリティがない
理論屋さんはOKでもDB屋さんとしては困る

VCFの breakend 記法で書かれているような、切れたもう一方の端がどこに繋がっているか分からないものがありえる？ → gnomAD SV の CTX/BND には CHM2 も書いてあるのでOKだが、とくにBNDでCHM2がないものはある？

path は step のリストで、step は向きを含めた node、、nodeはリファレンス上のpositionを持つ

### Plan A

Starting from a genome grpah which already contains SVs to be annotated.

* Pros
* easy to obtain nodes before and after the node that are required to define paths (usually ommitted in the VCF thus needs to be complemented)

* Cons
* representation can be complicated when simple variants are located on the SV nodes. A node can have SNPs after the dupulication e.g., in case of CNV or dDUP-iDEL.

### Plan B

Adding new submissions of SVs to the genome graph.

* Pros
* practically more feasible as new submissions will usually report new SVs

* Cons
* nodes in a genome graph needs to be edited (break into two or more nodes etc.)

### Application to gnomAD SVs

### Application for de novo SVs

### GVO (ver. AshinoCore)

## Discussion

ゲノムグラフで表現できるSVはOK

世の中のVCFやサブミッションにはゲノムグラフにできないものもあるんじゃ？
（ファジーポジション）

### Can GFA represent any types of SVs?

Theoritically yes (really?)
* practically, each node can have SNVs inside... etc.
* see limitations below

Minimal requirements:

### SPARQL query of SNV/MNVs and SVs

How many types?

What's are the biological mechanisms to cause SVs?

### Interoperability of submissions

Same SV = Same RDF

limitations:
* left shift (left alignment)
* split node
* node1 -> node1a, node1b (then delete node1?)
* fuzzy position (dbVar, gnomAD VCF)
* might not be represented in genome graph?
* inter chromosomal / translocation
* might be OK if genome graph contains all chromosomes
* diploids? / trisomy? / cancer?

#### TogoVar, MGeND, JVar

TODOs

## Acknowledgements

* Jerven Bolleman
* Simon Heumos
* Dmitry Repchevsky

## References