Using Big Data to Treat the Whole Patient
- by 7wData
Big data and computational modeling have the potential to fundamentally change the way patients are diagnosed and treated. Since the 20th century, diseases and other biological systems have been considered in isolation—which is described as a reductionistic approach.
Today, interdisciplinary researchers are using the wealth of medical data being generated by new technology to take a more holistic approach to healthcare—focusing on biological processes within the patient. Instead of understanding complex diseases by studying their individual components, systems medicine looks at how cells, genes, and tissues interact to cause disease.
Since the term “systems medicine” was first coined in 1992, there’s been an explosion of interest in the field, with the number of papers published increasing by an estimated 41% per year up to 2015. Speakers at the 2nd Conference of the European Association of Systems Medicine 2018, held recently in Utrecht, the Netherlands, presented new computational models for studying complex diseases and assessing patient risk, as well as novel interdisciplinary research.
Understanding complex diseases on a molecular level is among the research topics covered by systems medicine. Complex diseases, such as asthma, Parkinson’s, multiple sclerosis, and cancer, are not caused by a single genetic mutation. Instead, multiple genes, signaling pathways, and regulatory networks are involved.
“That’s why it’s called network or systems medicine,” said Jan Baumbach, Ph.D., chair of experimental bioinformatics at the Technical University of Munich. “We’re not looking at individual gene sets. We’re using big data to look at these systems.”
One challenge is identifying the combinations of genes involved in a complex disease, Dr. Baumbach explained. Even in monogenetic diseases, such as Huntington’s disease, which are caused by a single gene mutation, the disease-causing gene is often expressed the same in patients with and without the disease. In more complex diseases, such as cancer, where multiple genes are mutated, the situation is even more difficult.
“If you have 22,000 genes, then you can check every gene individually, perhaps check all possible triplets, quadruplets, etc., but the number of possible combinations grows exponentially,” he explained. “We call this ‘combinatorial explosion.’ It’s totally impractical to compute the statistics for all possible combinations, even if your computer was the size of the sun, so you need clever computational techniques to take shortcuts—evaluating the most likely solutions first.”
Dr. Baumbach specializes in using artificial intelligence (AI) approaches, such as machine learning and random forest, to find the most likely combinations of genes. His trick is to pick ensembles of genes that, although not individually linked to a disease, are part of the same molecular pathways. In patients, these groups of genes can work together to drive a disease state.
He also uses unsupervised learning to stratify patients with, for instance, asthma into subgroups from scratch (called de novo endophenotyping) using disease mechanisms, rather than individual genes or gene panels. With this technique, patients are deliberately not grouped by phenotype, but are—instead—clustered in subgroups by an AI.
According to Dr. Baumbach, “If you try a new medicine on 10,000 asthma patients, it might be very effective in 20%, but not effective in the other 80%, so there’s no chance of getting this drug registered, unless you have an effective means of telling the 20% from the 80%. If we can stratify these individuals into subgroups that share common mechanisms, we can check the mechanism against the drug target and look at other treatment options.”
Dr.
[Social9_Share class=”s9-widget-wrapper”]
Upcoming Events
From Text to Value: Pairing Text Analytics and Generative AI
21 May 2024
5 PM CET – 6 PM CET
Read More