Fujitsu & Kyoto University Create AI DNA Disorder Detection System
20-03-2021 | By Robin Mitchell
Recently, Fujitsu and Kyoto University developed a system that uses AI to detect genetic disorders in DNA samples. What challenges does sequencing DNA present, what has Fujitsu developed, and does such technology carry ethical issues?
What challenges does the human genome present?
The discovery of DNA sent shock-waves in the scientific community, and while the human genome has been mapped, exactly what each gene does is still undetermined. The mean length of a DNA strand in humans is 206.62cm, and each cell has 23 pairs of DNA strands. Furthermore, the total number of base pairs for the whole human genome (i.e. bits of information) is over 3,000,000,000. It is believed that scientists only truly understand less than 2% of the total sequence as most DNA appears to have no function.
Some genetic diseases can be easily identified as they either include a specific pattern of bases, or omit patterns. For example, albinism is a genetic condition whereby the human body does not produce melanin, resulting from a faulty OCA1 gene. Identifying such conditions can be done by sequencing an individual's DNA and looking for incorrect sequences.
Such comparative DNA sequencing is only possible thanks to the fact that all human DNA is almost identical. While features such as hair colour and eye colour vary, features such as organ growth, proteins, and enzymes do not vary.
However, identifying potential issues in DNA sequences can be tricky as scientists still do not understand how 2% of DNA interacts with the other 98% which appears to have no function. Still, it may in fact play an important role. Trying to look at patterns between patients and compare such long DNA strands to look for links is a task far too complex for the human brain.
Fujitsu and Kyoto University Create AI DNA Genetic Detector
With DNA being a huge and complex molecule, trying to understand how it interacts with itself is no task for people, but perfect for AI. As such, Fujitsu and Kyoto University have partnered together to create an AI system called MGeND, which can analyse DNA and genetic conditions to predict genetic disorders in patients.
When the patients DNA is entered into the system, it generates a result that includes the genetic disorder estimation and explanatory text as to why it has come to that conclusion. Many AI systems merely produce a result with no explanation. The system developed by Fujitsu and Kyoto University marks one of the first practical AI systems that can provide users with a reason to its conclusion.
The ability to produce an explanation also helps genetic researchers understand DNA sequences better that may have been overlooked previously. At the same time, doctors can benefit from such explanations when determining the course of treatment. As such, the AI system may become a critical tool for identifying cancers earlier, provide families with DNA analysis before having children, and offer better genomic therapies.
What ethical concerns are there?
As DNA sequencing and fixing become more accepted by society, there is a genuine danger of eugenics becoming normalised. Tools such as those developed by Fujitsu provide doctors with a great identification tool. Still, they could encourage its use to identify those who carry unwanted genes and prevent their spread into new generations.
Some would say that removing all bad genes is a good thing, for example, eliminating heart disease and strokes. However, once the medically bad genes are removed, some may consider other genes such as short stature, large or thin bodies, and hair loos to also be undesirable. Eventually, society would move towards genetic editing whereby humans can freely identify genes they don’t like and have them eliminated.
The technology developed by Fujitsu and Kyoto University, realistically, does not present ethical issues and instead helps researchers better understand the DNA genome. AI is highly optimised for looking at large amounts of data that is generally considered junk, only to find a pattern hidden from sight.
98% of DNA may be useless, but realistically, it is more likely that 98% of DNA is simply misunderstood. Hopefully, the system developed by Fujitsu and Kyoto University will continue to shed light on the human genome.
Read More
