Diving deep into data to crack the gene code on disease

February 14, 2014 by Karin Verspoor, The Conversation
The disease is written in our DNA code… somewhere. Credit: www.shutterstock.aom

The key to understanding disease is in our DNA, or the human genome which contains the instructions on how our body should develop and grow.

The key to progress in genomics research is in combining as much evidence from as many different people as possible, to sort out which DNA differences (or genetic variants) are connected to particular diseases, and which are just individual variation.

We can search or mine the published biomedical literature to help scientists find this evidence, and to help interpret the significance of any given genetic change.

But in a study released this week, my colleague Antonio Jimeno Yepes and I identified a new challenge for finding evidence about genetic variants in the mass of published literature.

Much of the information is available in PubMed, the biomedical journal citation database of the National Library of Medicine in the United States and the primary repository of biomedical research.

The PubMed search tool allows researchers to search through the abstracts of published articles, and the PubMed Central resource broadens that search to full text articles, albeit for a fraction of the overall literature (2.9-million of the 22-million articles in PubMed).

So far no biomedical search tool indexes what is known as the supplementary material of journal articles, extra data associated to a paper but not part of its narrative content. It turns out this is where the vast majority of detailed evidence about individual genetic changes is located.

Lots of small genetic changes determine who we are

Take any two people, sequence their DNA, and you will typically find that they are about 99.9% identical. The remaining 0.1% determines what makes a person unique, from the shape of their ear to their risk of getting sick with a particular disease.

What seems like an insignificant amount of difference translates to millions of tiny variations at the genomic level – variations in individual bases, or the 'A', 'C', 'T', and 'G' molecules that are the building blocks of DNA.

There can be simple substitutions of one base for another, called SNPs (Singular Nucleotide Polymorphisms, pronounced "snips"), or places where one or more bases are added (insertions) or removed (deletions).

Entire segments of DNA might be repeated or moved from one place to another. Some of these changes will have no effect, and some of them will have big effects.

Sorting out which is which, and especially which changes are related to disease and how, is a major focus of modern . Such research is possible at a level of detail and at a scale that was unimaginable not too long ago, thanks to recent improvements in DNA analysis technology called high-throughput sequencing.

Better diagnosis and treatment

Knowing which genetic changes contribute to disease risk will improve a doctor's ability to diagnose disease. It will also potentially help them to select the best treatment options, or predict the severity of the disease for a patient based on their genetic profile.

It will also lead to development of new treatments for the disease based on a deeper understanding of the underlying biology. The role of in explaining disease is hugely important information, with the power to improve patient health.

The basics of DNA. Credit: Shutterstock

The trouble is, those millions of tiny changes. Multiply that across even tens of people and you've got a whole heap of differences to sort through. The scale is daunting.

When you consider that it is not likely that any single change on its own will explain a person's disease risk, but rather many small changes acting together, you face an explosion of interaction combinations to explore. Biomedical researchers that are trying to make sense of all this data need some clues about where to start looking.

Finding nuggets of wisdom from the crowd

Luckily, there are plenty of researchers tackling these questions all across the globe, studying the genomes of groups of people with specific diseases and comparing them to healthy people.

When these studies identify important relationships between genetic variants and diseases, the results are usually published in a journal article. So when a researcher or clinician finds a genetic change that he or she suspects is associated with a disease, it is very possible that someone has already published some evidence that can help him confirm that suspicion.

Genome researchers spend huge amounts of time reading research publications, looking for such evidence.

But finding this evidence is hard, firstly because of just how many research publications there are to dig through. There are more than 22 million articles indexed in PubMed, and nearly a million new articles were added in 2013 alone.

Secondly, the way authors of these articles refer to genetic variants can vary tremendously. Some consistently follow the recommended nomenclature of the Human Genome Variation Society. Some authors use database identifiers, such as those from dbSNP, a large database of human SNPs. Others use more conversational ways to describe the variants, like "an adenine deletion in the mtrR promoter" [PubMed ID 23036167].

Automating the search for variants

Text mining tools have been developed to automatically identify mentions of genetic variants in the published literature. These include the Extractor of Mutations (EMU) and tmVar.

Such tools have been shown to work well; they compensate for much of the different ways variants are described in text and can normalise those descriptions to the standard nomenclature.

Our recent study in the journal Database, The Journal of Biological Databases and Curation applies such tools to PubMed abstracts and PubMed Central full text articles.

We aim to recover mentions of genetic variants of biological significance that have been captured in the Catalogue of Somatic Mutations in Cancer (COSMIC) database, at the Wellcome Trust Sanger Institute in the UK, and the International Society for Gastrointestinal Hereditary Tumours (InSiGHT).

We were expecting this to be a straightforward demonstration of the practical usefulness of text mining to help variant database curators and genome researchers.

We were surprised then to find that the tools only recovered less than 8% of the variants, even when we knew exactly which article to look at.

Our investigation led us to the conclusion that the vast majority of the information about genetic variants is being pulled from extra files associated with the publications, the supplementary material. When we processed these files with EMU, we were able to find more than 50% of the variants.

Broadening the search for evidence

Our study shows that the effectiveness of automatic literature mining methods for finding information about genetic variants hinges on access to the supplementary material.

This material is unfortunately not systematically indexed. Every publisher has a different way of storing and linking to the additional data files from publications. This diversity makes it difficult for automated tools to locate the data for processing.

Even if we could find all the relevant data files, there remain challenges.

When we considered the supplementary material, the EMU tool still missed about 50% of the curated variants. These are mostly variants that are included in tables or supplementary material in a way that is different from how they are typically referred to in text, such as when the information is spread across columns in a spreadsheet.

Our usual text mining tools don't work well for data expressed in this way, so new strategies are needed to automatically extract and normalise the data presented in tabular form.

Finding the genetic variants in the published literature is only the starting point for the important work of understanding the biological role of those variants, and how they are connected to a disease. The results in those publications need to be interpreted and synthesised.

But given how many publications are coming out every day, the lack of consistency in how genetic variants are described, and the fact that the information often isn't even in the papers themselves, it is clear that genome researchers need effective text mining technology to help them reach that starting point.

We are still working to get them there.

Explore further: New method developed for ranking disease-causal mutations within whole genome sequences

Related Stories

New method developed for ranking disease-causal mutations within whole genome sequences

February 7, 2014
Researchers from the University of Washington and the HudsonAlpha Institute for Biotechnology have developed a new method for organizing and prioritizing genetic data. The Combined Annotation–Dependent Depletion, or CADD, ...

Why is type 2 diabetes an increasing problem?

January 9, 2014
Contrary to a common belief, researchers have shown that genetic regions associated with increased risk of type 2 diabetes were unlikely to have been beneficial to people at stages through human evolution.

Discovery may help to explain mystery of 'missing' genetic risk

February 13, 2014
A new study could help to answer an important riddle in our understanding of genetics: why research to look for the genetic causes of common diseases has failed to explain more than a fraction of the heritable risk of developing ...

Sieving through 'junk' DNA reveals cancer-causing genetic mutations

October 3, 2013
Researchers can now identify DNA regions within non-coding DNA, the major part of the genome that is not translated into a protein, where mutations can cause diseases such as cancer.

Seven new genetic regions linked to type 2 diabetes

February 9, 2014
Seven new genetic regions associated with type 2 diabetes have been identified in the largest study to date of the genetic basis of the disease.

Scientists develop new approach to study how genetic variants affect gene expression

January 10, 2014
(Medical Xpress)—Each individual carries a unique version of the human genome. Genetic differences can influence traits such as height, weight and vulnerability to disease, but precisely what these genetic variants are ...

Recommended for you

Scientists provide insight into genetic basis of neuropsychiatric disorders

July 21, 2017
A study by scientists at the Children's Medical Center Research Institute at UT Southwestern (CRI) is providing insight into the genetic basis of neuropsychiatric disorders. In this research, the first mouse model of a mutation ...

Scientists identify new way cells turn off genes

July 19, 2017
Cells have more than one trick up their sleeve for controlling certain genes that regulate fetal growth and development.

South Asian genomes could be boon for disease research, scientists say

July 18, 2017
The Indian subcontinent's massive population is nearing 1.5 billion according to recent accounts. But that population is far from monolithic; it's made up of nearly 5,000 well-defined sub-groups, making the region one of ...

Mutant yeast reveals details of the aberrant genomic machinery of children's high-grade gliomas

July 18, 2017
St. Jude Children's Research Hospital biologists have used engineered yeast cells to discover how a mutation that is frequently found in pediatric brain tumor high-grade glioma triggers a cascade of genomic malfunctions.

Late-breaking mutations may play an important role in autism

July 17, 2017
A study of nearly 6,000 families, combining three genetic sequencing technologies, finds that mutations that occur after conception play an important role in autism. A team led by investigators at Boston Children's Hospital ...

Newly identified genetic marker may help detect high-risk flu patients

July 17, 2017
Researchers have discovered an inherited genetic variation that may help identify patients at elevated risk for severe, potentially fatal influenza infections. The scientists have also linked the gene variant to a mechanism ...

0 comments

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.