Scientists are using machine learning to interpret “dark matter” DNA

Scientists at Gladstone Institutes are using machine learning to target genetic disorders in so-called genomic “dark matter”.

The computational method being used, called TargetFinder, predicts where non-coding DNA – the DNA that does not code for proteins – interacts with genes. By analysing big data, researchers are abble to connect mutations in genomic “dark matter” with the genes they affect, potentially revealing new targets for genetic disorders.

In the study, published in Nature Genetics, the team from Gladstone Institutes looked at fragments of non-coding DNA called enhancers which act like an instruction manual for a gene, dictating when and where a gene is turned on.

“Most genetic mutations that are associated with disease occur in enhancers, making them an incredibly important area of study,” said the study’s senior author, Katherine Pollard. “Before now, we struggled to understand how enhancers find the distant genes they act upon.”

dna

The new study revealed that, on a strand of DNA, enhancers can be millions of letters away from the gene they influence.

However, using machine learning technology, the researchers were able to analyse hundreds of existing datasets to look for patterns in the genome and identify where a gene and enhancer interact.

They discovered that when an enhancer is far away from the gene it affects, the two connect by forming a three-dimensional loop, like a bow on the genome.

“It’s remarkable that we can predict complex three-dimensional interactions from relatively simple data,” said biostatistician at Gladstone, Sean Whalen. “No one had looked at the information stored on loops before, and we were surprised to discover how important that information is.”

dna1

The new computational approach is a much cheaper and a less time-consuming way to identify gene-enhancer connections in the genome as performing experiments in the can take millions of dollars and years of research.

The technology also gives an insight into how DNA loops form and how they might break in disease.

“Our ability to predict the gene targets of enhancers so accurately enables us to link mutations in enhancers to the genes they target,” said Pollard. “Having that link is the first step towards using these connections to treat diseases.”

Gladstone is set to offer all of the code and data from TargetFinder online for free.

Scientists, software developers and artists have begun using VR to visualise genes and predict disease

A group of scientists, software developers and artists have taken to using virtual reality (VR) technology to visualise complex interactions between genes and their regulatory elements.

The team, which comprises of members from Oxford University, Universita’ di Napoli and Goldsmiths, University of London, have been using VR to visualise simulations of a composite of data from genome sequencing, data on the interactions of DNA and microscopy data.

When all this data is combined the team are provided with an interactive, 3D image that shows where different regions of the genome sit relative to others, and how they interact with each other.

“Being able to visualise such data is important because the human brain is very good at pattern recognition – we tend to think visually,” said Stephen Taylor, head of the Computational Biology Research Group at Oxford’s MRC Weatherall Institute of Molecular Medicine (WIMM).

“It began at a conference back in 2014 when we saw a demonstration by researchers from Goldsmiths who had used software called CSynth to model proteins in three dimensions. We began working with them, feeding in seemingly incomprehensible information derived from our studies of the human alpha globin gene cluster and we were amazed that what we saw on the screen was an instantly recognisable model.”

The team believe that being able to visualise the interactions between genes and their regulatory elements will allow them to understand the basis of human genetic diseases, and are currently applying their techniques to study genetic diseases such as diabetes, cancer and multiple sclerosis.

“Our ultimate aim in this area is to correct the faulty gene or its regulatory elements and be able to re-introduce the corrected cells into a patient’s bone marrow: to perfect this we have to fully understand how genes and their regulatory elements interact with one another” said Professor Doug Higgs, a principal researcher at the WIMM.

“Having virtual reality tools like this will enable researchers to efficiently combine their data to gain a much broader understanding of how the organisation of the genome affects gene expression, and how mutations and variants affect such interactions.”

There are around 37 trillion cells in the average adult human body, and each cell contains two meters of DNA tightly packed into its nucleus.

While the technology to sequence genomes is well established, it has been shown that the manner in which DNA is folded within each cell affects how genes are expressed.

“There are more than three billion base pairs in the human genome, and a change in just one of these can cause a problem. As a model we’ve been looking at the human alpha globin gene cluster to understand how variants in genes and their regulatory elements may cause human genetic disease,” said Prof Jim Hughes, associate professor of Genome Biology at Oxford University.

Using CRISPR, UK scientists edit DNA of human embryos

For the first time in the UK, scientists have altered human embryos. Using the gene-editing tool CRISPR, the scientists turned off the protein OCT4, which is thought to be important in early embryo development. In doing so, cells that normally go on to form the placenta, yolk sac and foetus failed to develop.

Source: BBC

Tesla and AMD developing AI chip for self-driving cars

Tesla has partnered with AMD to develop a dedicated chip that will handle autonomous driving tasks in its cars. Tesla's Autopilot programme is currently headed by former AMD chip architect Jim Keller, and it is said that more than 50 people are working on the initiative under his leadership.

Source: CNBC

Synthetic muscle developed that can lift 1,000 times its own weight

Scientists have used a 3D printing technique to create an artificial muscle that can lift 1,000 times its own weight. "It can push, pull, bend, twist, and lift weight. It's the closest artificial material equivalent we have to a natural muscle," said Dr Aslan Miriyev, from the Creative Machines lab.

Source: Telegraph

Head of AI at Google criticises "AI apocalypse" scaremongering

John Giannandrea, the senior vice president of engineering at Google, has condemned AI scaremongering, promoted by people like Elon Musk ."I just object to the hype and the sort of sound bites that some people have been making," said Giannandrea."I am definitely not worried about the AI apocalypse."

Source: CNBC

Scientists engineer antibody that attacks 99% of HIV strains

Scientists have engineered an antibody that attacks 99% of HIV strains and is built to attack three critical parts of the virus, which makes it harder for the HIV virus to resist its effects. The International Aids Society said it was an "exciting breakthrough". Human trials will begin in 2018.

Source: BBC

Facebook has a plan to stop fake news from influencing elections

Mark Zuckerberg has outlined nine steps that Facebook will take to "protect election integrity". “I care deeply about the democratic process and protecting its integrity," he said during a live broadcast on his Facebook page. "I don’t want anyone to use our tools to undermine our democracy.”