Read in PDF
Dr. Mira Marcus-Kalish, is currently serving as the Director of International Research Affairs at Tel Aviv University. We caught up with her last week and spoke about the data challenges in Precision Medicine. She also filled us in on her current projects,
"We are involved in numerous research projects including the European Human Brain Project in which I serve as the Vice Chair of the medical informatics sub-project. In the United States, we are part of the CHMPR NSF Center, working in several directions such as healthy aging, Alzheimer's and Parkinson's diseases, wearable sensors, etc. Other EU projects I work on include new leads for atherosclerosis, innovative photonic bio-sensors for personalized diagnosis, and therapy monitoring for genitourinary cancers."
Join Dr. Marcus-Kalish in Berlin where she will be speaking on:
Big vs. Small Data Analysis: Targeting "Disease Signatures" towards Precision Healthcare
- Be sure to look at the "big picture" as often as you can, on the micro- and macro-environmental features of the human body functioning.
- Know the challenges: The number of patients/ samples needed, the best fitting analysis tools and thresholds to justify your precise target according to the current statistical and mathematical needs.
- Find different ways, targeted tools and methodologies to ensure reliability, and replicability.
- In the pre-processing phase, the cleaning, imputation and the curation of the data and the missing value must be taken care of before using the analysis tool.
Q: What are the controversies in Precision Medicine today?
First, there are many terms you can use: precision medicine, personalized medicine, precise medicine, and they are all basically targeting the same thing. These terms mean trying to provide the patient with the best tailored treatment and therapy. In my opinion, these are still buzzwords, due to the fact that we are not genuinely there yet. We have made many advances, especially in genetics, but we cannot really "prescribe" personalized medicine yet.
Precision medicine means what truly fits a specific disease in the best way. We are on the correct path: We know that the physical and mental environment matter a lot - the pollution in the air, the culture, the family situation, the environment - and all of these factors are beyond the biology and beyond the genetics. Trying to treat all of these parameters, and to look at the big picture of a specific individual, any human being for that matter, and observe them functioning in their surroundings, is not a place we have arrived at yet.
Q: Why aren't we there yet?
Our approach in the development of medicine, science and healthcare has come a long way. Originally, it was structured to investigate different areas separately and vertically. Maybe it was meant to be so, otherwise, we might not have advanced as far as we did. Today, we have reached a mature stage where we combine all of this data and knowledge, and look at the big scale picture. The challenge is not only the data, though. We have very good tools to analyze big data, but in order to treat a given person in the best possible way, we need sophisticated tools to analyze small amounts of data in a reliable and predictive manner. The small data is a number of people with exactly the same profile, which usually equals a small number. The statistical and mathematical tools are not there yet, even though we are working hard on their development. We need to combine our knowledge, and expertise, as well as the experience with all collected data, for example, to fine tune our analyses.
Further, there is a major controversy, nowadays, in scientific publications focusing on replicability, reproducibility, and liability of the experiment results, clinical trials, and thus the suggested treatments, drugs, etc. Therefore, many groups are developing tools to ensure the reliability and replicability of clinical trial outcomes. For example, there have been tools developed for dealing with the data, including for missing values, imputation, etc; simulation tools ensure the best fitting analysis tools, as well as hierarchical tools, and selective inference using FDR (False Discovery Rate) etc. All of these tools are needed to ensure that what we are providing is strong and trustworthy. It is critical that another scientist can repeat and verify our experiments.
Q: How will you challenge the participants during your presentation in Berlin?
It is my belief that just talking about things by slogans and not giving examples, will not help us. I will show both the challenges and how we can overcome them in different examples.
The first barrier with any data, especially in hospital data, but also in cohorts, like the ADNI-American Alzheimer's data bank, is for example, the missing values. The missing values have a major impact on the imputation of the data and the curation of data.
I will be presenting examples from the Human Brain Project on Alzheimer's and on Parkinson's diseases. We developed a new tool combining knowledge and data, using the ADNI. Even though many papers were already published using that data, there are still challenges. Another cohort is the hospital cohort on Parkinson's disease, which was collected at a major Tel Aviv hospital over the last 5 years. The two examples highlight the genetic correlations between Parkinson's disease and identifies new subtypes in Alzheimer's disease.
Q: What are the next steps to overcome the challenges?
Two major things. First, look at the big picture, the "broadband". For any disease, any patient, any human being, one must look at all the information that is available and relevant. What is the area they live in - the physical and mental environment, his or her history, the family, the biology? In summary, what is the micro- and macro-environment are they functioning in?
Secondly, sometimes you won't have enough information or the best tools. Once you have collected as much data as possible, use the best refined and reliable tools. Ensure that any prediction you make will be as reliable as possible.
Read in PDF
Join us at Controversies in Precision Medicine in Berlin, 13-15 November and hear Dr. Marcus-Kalish speak on: Big vs. Small Data Analysis: Targeting "Disease Signatures" towards Precision Healthcare.
Be part of 150+ senior level attendees from leading industry companies, academia and government institutions who are ready to address the challenges to ignite solutions.