TCGC: Clinical Genome ConferenceTCGC: Clinical Genome ConferenceTCGC: Clinical Genome Conference

Breast Cancer Genomics, Predictive Data and Therapy Guidance

Click Here to Return to Download Page  

Ann Nguyen:

Hi there. Welcome to this podcast from Cambridge Healthtech Institute for 2015's TCGC: The Clinical Genome Conference, taking place June 22-24 in San Francisco, California. I'm Ann Nguyen, the event's associate conference producer.

We're so pleased to have Dr. Laura van 't Veer with us today. She's the director of Applied Genomics and the Angela and Shu Kai Chan Endowed Chair for the Helen Diller Family Comprehensive Cancer Center at the University of California, San Francisco. She's also a keynote presenter at the conference, speaking about “Big Data in Translational Cancer Genomics”.

Laura, welcome and thank you for your time.

Laura van 't Veer:

Thank you very much, and my pleasure.

Ann Nguyen:

Your research centers on personalized medicine based on the genetic makeup of tumors and of patients as a whole. Why that area of interest, and what kind of resources do you have at UCSF for supporting it?

Laura van 't Veer:

I think my interest into looking into the patient by the genetic makeup of the tumors as well as the patient as a whole is in that the last 20-25 years of molecular biology research has taught us that every person is different, as we all know, but that actually the genes that everybody has are different, and that in some of those there's information that can foretell the risk of developing cancer.

My area of study is mostly breast cancer, so there's some of these genes that a person can carry that is different from another one, and in some cases these are so-called mutations in breast cancer genes, breast cancer BRCA1 and BRCA2, that actually confer a high risk to develop breast cancer.

So my interest for the last 20 years has also been to understand why in some families breast cancer occurs more frequently than average in the population and to use that information to guide more intensified screening processes or preventive interventions that we might have available but also to use that information even in the treatment of cancer if in such individual breast cancer may occur, because we know nowadays that there might be for a person that is a carrier of a BRCA1 gene mutation, even if that person then develops breast cancer, we have more specified therapies that are very effective on the tumors that arise in those women.

So that brings me to the second part of my interest, which is the genetic makeup of tumors. Tumor cells are only very slightly different from normal cells in the body, in that these normal cells have acquired, during life, gene mutations and those changes in the genes in the normal cells actually confer that the normal cell starts to grow more fast than it should be, and thereby becomes a cancerous process. So for different tumor types, we actually know a number of what we call the drivers of that accelerated cell division, so the accelerated growth of that mass…for instance, in the breast when a tumor arises there is a tumor mass, and the tumor mass arises because normal cells start to divide more quickly. Besides that we can identify what are the changes in the genes between the normal cells and in my case often the breast tumor cells. We can then again use that information to actually guide the choice of a therapy because over the last 20 years or so, more and more specific so-called targeted therapies have been developed that specifically would block the growth of a tumor, so the activity of a tumor cell, by directly blocking that one change in a gene that we identify by molecular tests.

And so at UCSF here in San Francisco, we are matching the treatment for a breast cancer patient specific to the biological changes that we identify in the tumor that that woman has. We follow whether, if we choose a treatment where we think the right match is found, then we monitor whether we are right in our predictions and we want to learn from looking at every individual patient in the study if our prediction is correct, and if not, why not, so that for the next patient that comes in the trial, we can adapt our predictions and choose a more optimal treatment for a patient than we maybe initially would have predicted, and that's a large clinical trial called the I-SPY trial.

Ann Nguyen:

Your work has been quite transdisciplinary across the sciences, for instance, during a 2002 study that involved physicists, statisticians, computational scientists, medical oncologists, pathologists and biologists. How do you determine collaborators beyond the usual suspects and steer everybody toward the same goal of clinical utility and improved patient management?

Laura van 't Veer:

I think that's an interesting question and it's also an interesting experience. I must say in 2002, which is some time ago, I was often very overwhelmed with the knowledge of people in all these other areas of expertise like physicists, statisticians, computational scientists, because of course everyone is an expert in their own area of study, in their own area for expertise.

But as I explained to you where, what we call personalized medicine, if we personalize the diagnostic process and the treatments of one patient based on a more in-depth evaluation of the disease by looking at the molecular changes, the gene changes in a tumor or in a person, it means that you need to study how that person's disease compares with a large reference of other individuals with similar but yet a little different disease. So to be able to do that you need to use computational methods, so comparison methods between that one person as part of a large group, to understand which are the differences that are essential and really drive the disease in that particular person.

You can approach such a problem from a pure mathematical approach in which you can reach a result that would be statistically optimal. However, if you approach the problem where you want to make sure that you don't make any -- or you do not diagnose somebody who has a disease, because your methods will say that person does not show the alteration that I associate with disease, so-called "false negatives" -- if you develop a method that would have many false negatives, that would mean that you would miss to diagnose somebody with a disease or with a potential option for a treatment, because your methods might be statistically most optimal but not most optimal from a patient perspective. Because if you are a patient, you want to be the one that is diagnosed in the right way.

What I actually did over the years, and it started and went to study in 2002, MammaPrint in breast cancer, is that I made the statisticians aware that the mathematical optimum is not always the clinical and patient optimum. That was actually an interesting process to make them learn that they're -- or make them aware (and also myself aware), that in different worlds you can achieve the best answer, but if you achieve the best answer from the mathematical point of view, that might not be the best answer from the patient's point of view.

So, what I've been doing is explaining to these mathematicians and statisticians that they need to think about analyzing and predicting what they're doing from a patient perspective. And actually everybody has been very appreciative that I brought that knowledge into the group of people. So the interactions with these people with other expertise was by chance, but I think over the years I've been able to find people to work with that are actually interested to sort of look over the boundaries of their own area of expertise.

Ann Nguyen:

You'll be speaking on June 22 about molecular genomic data integration and its relation to cancer patient therapy and survival prediction. What's the main theme you intend to convey?

Laura van 't Veer:

What I will discuss on June 22 is actually indeed by looking into the personalized medicine approach, which is to, when somebody's being diagnosed with cancer, to really look in depth in all the levels of molecular knowledge that we have available, at that person's tumor, in my case, breast cancer. So what we are evaluating is, for instance, by the test MammaPrint I developed, whether the tumor has a high risk for recurrence or a low risk for recurrence. If the risk for recurrence is very low, then very modest treatments, often hormonal treatment, is sufficient beyond surgery and local radial therapy to have a successful treatment of the patient.

However, if somebody is high risk of recurrence, our current standard therapeutics would be able to successfully treat that breast cancer in 30%-35% of the individuals. So the clinical trial I'm working on, I-SPY 2, in addition to the standard therapy, tests new, very specific targeted drugs to increase that percentage of women that respond completely to the therapeutic regimen from 30% to 60%.

So over the last four years, we've been working on this I-SPY 2 trial where we are testing new targeted drugs, and we have successfully, with a whole group of people of course, and patients that participated in the trial, we have found it for two drugs already, whom we need to give that particular type of drug to reach this 60% success of response. It's a so-called "master protocol", which means that we can test drugs sequentially, so if we have tested one drug then the next drug gets into the trial, and so we are very efficient in understanding who are the patients who respond to what therapeutic targeted drugs.

Ann Nguyen:

That all sounds great. Thank you so much for sharing a glimpse of your experiences and insights. You're obviously doing a lot of very important work, so we're looking forward to learning a lot more from you this summer.

That was Laura van 't Veer of UCSF. She'll be giving her keynote presentation during the opening session at TCGC: The Clinical Genome Conference, which runs June 22-24 in San Francisco. To learn more from her in person, visit for registration information and enter the keycode "Podcast".

This is Ann Nguyen. Thanks for listening.