Julián Isla's life took a tragic turn now ten years ago. Sergio, his young son, began to show at three months of age the first symptoms of a complex disease and doctors were not able to find the cause. Eight months later, it was learned that the boy had the Dravet syndrome, an epileptic encephalopathy that manifests itself in the first year of life.
Isla - who works as head of data consulting resources and artificial intelligence (AI) in Microsoft Spain - says that, “unfortunately, before being correctly diagnosed, Sergio was prescribed a medication that was contraindicated for the disease that has not yet It was known that he had. ”
After taking it, “he began to suffer dozens of epileptic seizures a day. The brain damage that this could cause is something we do not know, but what is certain is that we could have avoided it with an early diagnosis, ”he laments.
Before being correctly diagnosed, Sergio was prescribed a contraindicated medication that had devastating effects.
Rare disease patients “need approximately five to six years to have a proper diagnosis and visit up to seven specialists. And, despite everything, more than half remain undiagnosed, ”he emphasizes. A fact that corroborated Mark Caulfield, Project Manager 100,000 Genomes in British public health (NHS), in a recent interview with Sinc.
“The fundamental reason for these shortcomings is that until recently - or even today - the analysis of genetic information (80% of rare diseases are genetic) was an expensive and laborious process that biomedical professionals had to do manually in the genetic analysis laboratories, ”explains Isla.
Harness the power of the machines
So, in the midst of suffering and despair, this engineer wondered why, "in an era marked by technology, doctors were not using the power of computers and data in the diagnosis of rare diseases." And he began to think about what he personally could do to improve this situation.
“Of course, the first thing you try is to find a cure for your child's rare disease until you realize that it is a very complicated task that will take a long time. Then I asked myself how I could use the technological resources that I have as an engineer software from Microsoft to help create a computing platform that will use artificial intelligence and help change this scenario. ”
The question that was asked was “if AI could be used to solve a problem that humans have a hard time right now, which is to analyze the genome of a person, see the variants or mutations they may have and try to find the relationships between those mutations and the symptoms. There is a perception that machines are going to be able to do everything and it is not like that, but for intensive calculations and pattern searches they are very good. And the analysis of genetic information is a good field of application, ”he emphasizes.
Let's do it
To launch the project, Julián Isla brought together a group of men and women related to the world of technology and health and created in 2017 Foundation 29, a non-profit organization, which takes its name from the International Day of Rare Diseases, celebrated on February 29.
A team of employees and volunteers of this foundation, led by Isla, has been responsible for the design of Dx29, a tool based on artificial intelligence and created to support doctors in diagnosis.
As detailed by the engineer of software, “The process begins with the automatic identification and coding of symptoms from medical reports. Then Dx29 allows doctors to abstract from the complexity of gene identification, simply by selecting the symptoms that the artificial intelligence engine proposes as more likely. ”
In the final step, "once enough symptoms have been compared with genetic information, the tool presents a classified list of potential conditions for the doctor to evaluate and decide how to act." This whole process - he adds - "does not take more than ten minutes"
What Dx29 does is facilitate and automate genetic analysis. "Any doctor who is able to identify symptoms and describe them can use it," he says.
The platform - he explains - “consists of two fundamental components, one of them is the cloud computing that Microsoft has donated for the project. And then there is the technology of the algorithms themselves introduced in the analysis process, in whose development we have collaborated with several leading research centers in the world ”.
For example, “the symptom extraction system is based on an algorithm of deep learning,developed alongside Toronto Pediatric Hospital. We have made the gene prioritization tool with a consortium of researchers from Europe and the United States, called The Monarch Initiative. And for the algorithm that allows that prioritization when you don't have genetic information about the patient, we have worked with the University of California. ”
Efficiency of 80% in the beta phase
Isla says that Dx29 is already open to be tested in beta mode. In fact, he says, "several Spanish institutions, including La Paz Hospital in Madrid, have used it with patients already diagnosed and have shown an efficiency of 80%."
Now Isla and its foundation are preparing the clinical trials of the tool with the British NHS and with the public health systems of the Community of Madrid, the Basque Country and Catalonia. In total, it will prove its effectiveness with 600 patients already diagnosed.
He says that tests with patients also have another function, which is to help debug information about rare diseases. “There is a very worrying issue and it is that general databases - which are those used for genetic diagnostic tools - have been built over the years by technicians who were not doctors, even if they used medical documentation. And we have found many mistakes. ”
Therefore, he adds, “we want to use the information provided by patients and their families to allow us to profile rare diseases well. If each patient helps us identify the symptoms and characteristics of their disease, the tool will be more accurate. ”
Collaboration of patients and families
Dx29 is also available to patients, or their families, from June 7. "What they can do, those who are already diagnosed, is to help us define their phenotype - their symptoms - so that the platform learns and can help find patients like them faster."
This is fundamental “because if the AI part of data that is incorrect it will produce erroneous results. Right now there is too much data to clean from the point of view of disease description. And that is fantastic ground for patients to help us, ”he says.
But for the project to be viable, financing is needed, as always. Isla insists that the mission of Dx29 is to “break the barrier of the diagnosis of rare diseases and for that to happen we must democratize access. We want a primary care physician who does not have great means to use it. So we have to articulate the mechanisms so that the platform can be financed in the long term. So far, this has come forward only with our personal effort, ”he remarks.
"We are preparing clinical trials with 600 patients in hospitals of the British NHS and public health in Madrid, the Basque Country and Catalonia"
Julian Isla says that when his son became ill ten years ago “he was very naive and thought that things were like that, that the diagnosis was difficult and would take time. I was quite confident in the doctors' ability to do so, but I did not realize that they did not have adequate means to detect a rare condition. They were based on their own clinical eye, on their experience and in these cases it is often not enough. ”
That is why - he says - "we have dedicated so much effort to launch a platform like Dx29". I spent eight months without knowing what was happening to my son. If doctors had had this tool at that time, they could have known within minutes. ”
The battle from multiple fronts
Island, in addition to leading the design of this new tool, battles on multiple fronts to contribute to the best treatment of rare diseases. Among other things, he participates in the committee that gives the designation of an orphan drug to new therapies for these pathologies in the European Medicines Agency and is a member of the scientific advisory committee of the Center for Biomedical Research in Rare Disease Network(CIBERER).
He is also scientific director of the European Federation of Dravet Syndrome, an organization that groups patients and caregivers of people with this disease in sixteen European countries.
Sergio's current situation - he says - “is complicated, has a disease with a very severe symptomatology for what is normal in Dravet syndrome. It's hard. If I do all this, it is to avoid that other fathers and mothers have to go through the same for what I went through until we had the diagnosis. ”
When asked how he gives his life to do so many things, Julián Isla says that when you have a child with such a severe illness “the vital wound is so deep that sometimes the ability to give others is the only way to relieve That wound That's where the energy comes from. ”
The tool designed by Julián Isla and his team at Fundación 29 has been conceived as a support instrument for doctors to improve the diagnosis of rare diseases.
However, he believes that in the future this work will be done primarily by machines - artificial intelligence. “The genomic complexity and the variability of symptoms of these pathologies make the traditional medical classification and diagnosis system fall short. Therefore, we want to advance in the diagnosis of precision for these diseases, ”he emphasizes.
The engineer ventures that future patients “are probably not going to have a diagnosis with a name of the doctor who discovered it or the affected gene. They will have something like a numerical code - or coordinates - that place their ailment in a complex multidimensional space that reflects all angles of the disease. And we want to take the first steps in that direction, ”he concludes.
. (tagsToTranslate) father (t) want (t) improve (t) diagnosis (t) rare disease (t) (t) rare (t) artificial intelligence (t) (t) engineer (t) software (t) design ( t) platform (t) expedite (t) help (t) mother (t) pass (t) occur (t) child (t) be (t) diagnose