It’s early days for Big Data, according to Tracy Mayne, who heads Medical Affairs Strategic Research at Purdue Pharma. Perhaps this is why he compares it to teenagers’ fascination with sex.
“Everyone is talking about it, everyone thinks everyone else is doing it, but the truth is there are not very many people doing it, and there are few who are doing it well,” he said introducing the panel discussion entitled, “What Big Data Means For Bioinformatics,” at BIO-Europe Spring® in Barcelona.
Salvador Capella Gutiérrez agreed, saying that “everyone is talking about big data, but when you look you realize that there is nothing that is big about the big data people are talking about. There is the promise, but companies tend to oversell. Expectations are certainly not being met, and while I believe they will be met, it is difficult to predict when.”
Capella Gutiérrez is the Chief Technical Officer for Spain’s National Bioinformatics Institute, and also serves as Spain’s Technical Coordinator for the European Union’s ELIXIR program, which when expectations are met, will serve as a major component for Big Data analytics.
An initiative to build a single pan-European infrastructure to serve as a repository for all types of sequence and genotype experiments, ELIXIR has won participation from 21 Member States of the EU gathering over 180 research organizations.
There are different stages to this work, and the one that is closest to pharmaceutical is the European Genome-phenome Archive (EGA), he said, “which also is very much a big data thing with not only genomic but also proteomic data. Interestingly, some 25 percent of the data is coming from the pharmaceutical industry contributing native data from experiments.”
Alfons Nonell-Canals is the CEO at Mind the Byte, a bioinformatics company he founded in 2011 as a specialist in computational drug discovery using a pay-per-use software-as-a-service platform.
“We are at the very beginning of the drug discovery with an ability to make predictions about proteins. We manage huge amounts of data using a big data approach, but this cannot be called big data. Our aim, however is to integrate with big data in the process, and I have some ideas on how pharma could use the data.”
Athula Herath is the Global Head of Real World Evidence Disease Epidemiology for Novartis who said a large part of his work is to leverage large data sets to better understand diseases and patient experiences throughout a product’s lifecycle. As part of Medical Affairs at Novartis, he said the group’s goal is to establish the value proposition in economic terms, but also for the purpose of guidance and even guidelines for the medical community.
“The data that comes out of clinical trials and experiments is drawn from an artificial environment,” he said. “And it is a fair comment when physicians say that our data about drug effectiveness is not always relevant to what they see in the real world.
“Questions that can be addressed can be related to burden of the disease, drug utilization, safety, and effectiveness,” said Herath.
Across a fragmented landscape in Europe he seeks diverse sources with well-characterized outcomes to gain an ability to establish phenotypes and the state of a disease a physician is treating.
“One of the sources comes from the insurance industry, which traditionally collects great amounts of data. And in Europe there is long tradition of compiling cohort registries. In the case of the Karolinska Institute [Sweden] it stretches back 150 years. Another example of a source is in the United Kingdom where since 1989 some 4,000 general practitioners have collected their data providing longitudinal views on outcomes. More recently we have seen the creation of programs in the UK for 100,000 genomes and the BioBank.
This is a first stage of the work, using what I would call the good, disciplined big data to establish these phenotypes. Less good big data, he said, would include thousands of sources such as social media.
After this we address what would be called the need. In precision medicine what they are really talking about is what we would term patient stratification that allows the identification of a patient group much more precisely and establishes a benefit-risk ratio for these patients. This then allows the creation of what we call endotypes, a subtype of a condition with a distinct pathobiological mechanism. Now we would be looking into gene expression, whole genomes and other factors to very well characterize these patient sub populations.
In a sense this takes us back again to the beginning where we can understand what mechanisms drive the disease for this particular group and a target. This is very different for Novartis from an approach we see elsewhere in biotech where someone walks in the door and says they have a target and now they are looking for the disease. Perhaps that approach worked 25 years ago. It is not what will work now.
Rise to new heights at our next event, BioPharm America. We offer powerful opportunities for you to network with the right partners and bring together hundreds of the world’s most innovative leaders across biotech, finance and pharma for high-level networking, pre-scheduled partnering meetings, strategic panel discussions and more.