Preview Mode Links will not work in preview mode

Getting Personal: Omics of the Heart

Sep 27, 2017

Jane:                                  Hi everyone, welcome to episode five of Getting Personal: A Mix of the Harsh. I'm Jane Ferguson an assistant professor at Vanderbilt University Medical Center and chair of the FGTB Professional Education and Publications Committee. This month we start off by discussing a topic that isn't strictly scientific but may have just as big of an impact on your career as your science, mentoring. I'll talk to Anna Pilbrow from the FGTB Early Career Committee on how you can find the right mentor for you. And if you've been around long enough that you know longer need any more mentoring, keep listening as we would love you to sign up to become a mentor and share your wisdom with the next generation. Then I'll talk to Naveen Pereira from Mayo Clinic, about some of the papers we've been reading this month.

                                           So I'm here with Anna Pilbrow, who is a member of the FGTB Early Career Committee. So welcome Anna, and could you take a moment to introduce yourself?

Anna Pilbrow:                  Thanks Jane sure. So I'm a senior research fellow at the Christchurch Heart Institute at the University of Otago in Christchurch, New Zealand, and I'm really interested in trying to understand the mechanisms underlying inherited susceptibility to heart disease and also trying to find new biomarkers that predict incident cardiovascular events in asymptomatic people. And it's those interests really that lead me to the FGTB Council and to becoming involved in the Early Career Committee.

Jane:                                  Yeah that sounds really interesting and as part of your involvement in the FGTB Council I know you've been doing a lot of outreach and today we're here to talk about mentoring. So mentorship is something that I think we all recognize is really crucially important but sometimes the mentor-mentee relationship can fall short of people's expectations, it can subject both mentees and mentors to a lot of frustration. So is this something that you've been hearing from Early Career members?

Anna Pilbrow:                  Yes, yes. Unfortunately, stories of mentoring relationships going wrong is something we hear all too often. I should stress that there are many wonderful mentors out there, but there are also plenty of empty mentors and that's something that we all need to be aware of.

Jane:                                  Absolutely. You know this podcast is focused mostly on personalized medicine, but I think personalization and precision are things that we also need to apply to our careers. So I've heard the excellent advice that you should seek out multiple mentors, really as many as you can handle. I think most of us would never say, "I already have a friend. I don't need another one." You know but some people think that, "Oh I have a mentor, I'm good." But, I think even if you have a supervisor who does act as your mentor, which not all supervisors do also act as mentors, but even in that case I think it's still important to try to find other mentors who can offer different perspectives.

                                           You know even, very wise senior mentors are limited by their own experiences and their own implicit biases, they can never give you everything that you need and I think they shouldn't. That's not what a mentor should do. So in an ideal scenario, people I think would build up their own personalized network of mentors spread across different locations, who can each offer something unique on an as-needed basis. But, as a junior person it can be really intimidating to go up to someone and just ask them to mentor you, which I think it can be sort of like asking a complete stranger to be your new best friend, which isn't always the most comfortable interaction. But, are there any ways to make that process easier? So say you're early in your career and you'd like to find a mentor, how would you go about doing that?

Anna Pilbrow:                  Oh gosh, that's a great question. I completely agree with all you've said and this is exactly why the Early Career Committee has initiated the FGTB Mentoring Program. So as an Early Career member, all you have to do is sign up and we'll do the rest.

Jane:                                  Well that sounds fantastic. Can you explain more about how that process works and what's required?

Anna Pilbrow:                  Sure. So the aim of the FGTB Mentoring Program is to connect Early Career members within the council to a senior or a peer mentor also from within the council, and the senior and peer mentors will have expertise in the field of functional genomics and translational biology, and they'll also have expertise in the area that the individual want mentoring in. So if you're looking for a mentor, or you'd like to be a mentor, all you need to do is fill out a short form on our website. The full web address is really long and I actually find it easiest to get the add simply by Googling AHA FGTB Mentoring Program, or you can go to the FGTB council webpage, click on the early career tab and if you scroll all the way down to the bottom you'll get to a link that will take you to the Mentoring Program page. And so once you're on that page, you'll find the links to the mentee and mentor application pages.

                                           And one tip I have for Early Career members looking for a mentor, is to think quite carefully about the kind of mentoring that they want. Is it particular aspects of their career development? Things like grant writing, maybe applying for their first job? Or is it more technical things to do with a particular experimental design or something like that? And Early Career members can be really specific when they fill out that form on the website, that will help us match them with the best mentor that has the expertise in that field.

                                           And so once it's set up a match between a mentee and a mentor, what we would typically expect is that the mentoring relationship would last between sort of four to six months, and during that time we'd expect there to be regular contact between the mentor and the mentee, and that can be either by phone or email or some other electronic communication. Ideally, we'd really like the mentor and the mentee to make face-to-face at least one time during that four to six month period. So, a great way to do that is to meet up at a conference such as AHA Scientific Sessions and once the four to six months is completed, then we ask the mentor and the mentee to complete a short exit survey so we can get some feedback on how things have gone. I must say, I've been delighted with the positive feedback we've received so far with the program.

Jane:                                  That sounds really really great. So who do you think could benefit from taking part in this program?

Anna Pilbrow:                  Everybody. Absolutely everybody. So the advantage of this program is that mentees are individually matched with their mentors so that each match should uniquely address the requirements of that mentee. And so, because you can enter exactly what you're looking for when you sign up so you have the best chance of being matched with someone who can help. So one particularly unique aspect is that we have peer mentors as well as the senior mentors, and sometimes you know someone who's just a little bit ahead of where you are now can actually offer you really valuable advice and give a really neat perspective you know compared with somebody who is many years or decades ahead of you in their career. And the other thing I'd mention is that to encourage FGTB members who live outside of the U.S. to also apply to this program. So I'm based in New Zealand and sometimes that makes me feel a little bit isolated compared with colleagues in other countries around the world, and this program really is a great way you know for everybody to expand [inaudible 00:08:04] and become engaged with the council no matter where you are.

Jane:                                  I think that's a really great point and it sounds like a great program. So the peer mentoring thing is interesting to me as well. How would you know if you should sign up as a mentee or a peer mentor?

Anna Pilbrow:                  Oh that very much depends on what you need, and it's important to remember that you can actually do both. So, if you're still early in your career, you can still offer something to people who are just behind you and would really love to have you become a peer mentor. But that doesn't also mean that you can't be a mentee yourself and be matched with a senior mentor of your own.

Jane:                                  So is there a limit to how many mentors you can be matched with through the program?

Anna Pilbrow:                  Absolutely not. So we ask that mentors and mentees try the relationship out for four to six months and to try a face-to-face meeting at AHA Sessions in November for example or some other time. And if that doesn't work out or if you just want another mentor, you can sign up again the following year. So there's absolutely no limit to how many times you can sign up and additionally, if you're looking for mentorship in several distinct areas and need a few mentors simultaneously, that's fine as well. So just let us know that when you sign up and we'll try to find appropriate mentors for you.

Jane:                                  That sounds great. So I know registration for the AHA Sessions in November just opened up, but people are probably just starting to plan their trips. So what can they do now?

Anna Pilbrow:                  Right. It's a great time to start thinking about this. So if you're early in your career and you want general mentorship on navigating AHA Sessions or career planning or if you are looking for specific mentorship in a given topic area, sign up now to become a mentee. And if you're further along in your career and you've developed expertise that could be useful to others, it'd be great for people to sign up as a mentor. So, make sure that you thinking about Sessions coming up, make sure that you schedule time to meet with your mentor or mentee during the meeting. And also, this is a great time to also sign up and plan to attend the early career day which is held the day before the main meeting on Saturday November 10.

Jane:                                  So if when people have signed up, do you have any advice for people who are going through the program?

Anna Pilbrow:                  That's a great question. So, for both mentors and mentees it's really important to communicate throughout the process but particularly at the beginning to set expectations. So talk about how often you plan to meet, whether it's going to be by email, phone or in person, and be very clear about what you hope to gain from the relationship. And also, be nice. Like mentors tend to be really busy people so if your mentor doesn't respond to you right away, it's probably because they have a grant deadline or a pile of reviews to get through or a manuscript or back to back meetings or family things are going on. And we've all been in that position starting out, so having lots of questions and not knowing where to start this is all part of the normal process of being a mentee. So mentors need to keep that in mind as well and meet the mentee where they are. I guess it's all about being respectful of that relationship and being very clear about what you want to achieve.

Jane:                                  I think that's fantastic advice and of course, as members of the FGTB Council we can assure all prospective mentees that really everybody on the council is already very nice so we think that you won't have any problems being matched with some great mentees, great mentors and we really encourage people to sign up for this program. It's really valuable, you have nothing to lose and lot of potential things to gain from being part of this. So thank you so much for joining us Anna.

Anna Pilbrow:                  Thank you very much.

Jane:                                  Hi Naveen, how are you doing?

Naveen Pereira:              I'm doing well Jane. You know, I was reading "Circulation: Cardiovascular Genetics" and in the April issue of this year, there's a manuscript titled Non-familial Hypertrophic Cardiomyopathy: Prevalence, Natural History and Clinical Implications. The senior author on this paper is Chris Semsarian and he's from Australia. And essentially this manuscript highlights the fact that hypertrophic cardiomyopathy, which in large part is thought to be inherited, is also not inherited and perhaps it's important to differentiate the two phenotypes. And so they studied 413 patients coming to their clinic with hypertrophic cardiomyopathy and they found that 61% of these patients had no familial history and 40% of these patients had no sarcomeric mutations. And so, they deemed these patients to be a form of non-familial hypertrophic cardiomyopathy. These were older patients, males, patients with a history of hypertension and a non-asymmetrical septal morphology. What is important is that these non-familial type of hypertrophic cardiomyopathy patients usually have disease onset at the later stage in life and they also have less severe disease.

                                           So, I think when we try and prognosticate these patients and aim certain medical therapies towards these patients, we've got to consider whether they're familial or non-familial. And this work has also been highlighted before in the form of an article by Mike Ackerman and his group in Mayo, where they looked at the yield of genetic testing in hypertrophic cardiomyopathy. And essentially patients who are younger at the time of diagnosis, age less or equal to 45 years, patients who have severe left ventricle wall thickening greater or equal to 2 cm and patients who have familial history hypertrophic cardiomyopathy are more likely to have sarcomeric mutations than those who don't.

                                           So, both these papers kind of highlight the fact that we got to start thinking of hypertrophic cardiomyopathy as familial or inherited, or non-familial or non-inherited, because initially people thought, "Well you know, maybe we are missing mutations," but with whole genome sequencing, whole exome sequencing these patients with the non-familial hypertrophic cardiomyopathy tend not to have causative mutations. So I really wonder if it's a different disease entity from a molecular perspective.

Jane:                                  Yeah, that's really interesting and sort of raises the question of, "Is this non-genetic? Are there other genes? Is this sort of a multi-genic, poly-genic phenomenon where you know the just sort of whole exome sequencing or whole genome sequencing may not be able to identify the causal genes in a lot of these cases?" It's really interesting.

Naveen Pereira:              Right. And then, you know there was this other paper in Journal of American College of Cardiology that was published again in the April 4th issue, 2017, and it's titled "Autosomal Recessive Cardiomyopathy Presenting as Acute Myocarditis." And the senior authors are Bonnet, Gelb and Casanova. They shared equally senior authorship. And really, this paper addresses the issue as to why some children are predisposed to acute viral myocarditis, which can present fairly fulminantly, while some children don't despite a lot of kids having viral infections.

                                           And so, they tested the hypothesis that perhaps genetic variation in Toll-like Receptor 3 or in the interferon alpha beta immunity system can predispose these children to developing acute myocarditis. However, when they tested this hypothesis in vitro by using induced pluripotent stem cell-derived cardiomyocytes, looking at expression, looking at these genes, deficient cardiomyocytes, they weren't able to show a definite role as far as predisposition towards developing myocarditis for Toll-like Receptor or interferon. And what they surprisingly found was that 7 of the 42 patients that they studied by holding some sequencing that is about 17% of these patients, actually had a likely pathogenic mutations in six cardiomyopathy associated genes.

                                           So this raises the question overall that perhaps if you have genetic mutations in the cardiomyopathy associated genes, could you be predisposed to these cardiomyopathies or cardiovascular specific disorders, and should we be searching for mutations in these cardiomyopathy genes in other types of cardiomyopathy, like tachycardia-induced cardiomyopathy or Takotsubo's disease, etc.

Jane:                                  Yeah, that's really interesting. It's sort of a perfect example of a gene environment interaction where you know the genetic predisposition alone is not enough to cause disease, but then when combined with an environmental hit like a viral infection that's when the disease manifests. Very interesting.

Naveen Pereira:              Right. And I heard that you have found something interesting as far as machine learning is concerned, Jane?

Jane:                                  Yes, yes. So I was reading this paper, which was published in PLOS ONE in April. So the first and last authors of that paper are Steven Weng and Nadeem Qureshi. And the title of that paper was, "Can Machine Learning Improve Cardiovascular Risk Prediction Using Routine Clinical Data?" So, as the title suggests the authors were interested in whether they could improve on standard risk prediction algorithms by using an unbiased approach ... like machine learning approach. So they used the identified electronic health record data for over 350,000 people, and this was in the UK from UK Family Practices, and they took the baseline variables from people who were free of cardiovascular disease at the start of the study and then they looked to see if they could predict the risk of CVD events over the following 10 years.

                                           So they decided to compare four different machine learning approaches to see the efficacy of the different approaches and then they used the American College of Cardiology guidelines as the standard to compare these new computer approaches to. So in that model they included eight primary variables, which are included in the ACC/AHA algorithm such as age, sex, smoking, blood pressure, cholesterol and diabetes. And then for their machine learning algorithms, they added additional variables that were present in the EHR. So they had 22 variables in total and that included things like triglyceride, CRP, creatinine, also ethnicity and presence of other diseases such as rheumatoid arthritis or CKD, and then also whether there were certain prescribed medications.

                                           They took the first 75% of the sample as a training set, so that was over or just under 300,000 individuals. So they used this to train their various algorithms and then they used the remaining 25% just selected randomly, which was a little over 80,000 subjects, as a validation set to assess the efficacy of these algorithms. So over the 10 year period in total, 6.6% of the subjects developed incident CVD and they found that all of their four different algorithms, so that included random forest, logistic regression, gradient boosting machines and neural network approach, they all outperformed the established risk algorithms. So they all did slightly better than the current ACC guidelines. And the best one that they found was the neural network approach. So that correctly predicted 7.6% more patients who developed CVD compared with the established algorithm.

                                           So, it's really quite a significant improvement [inaudible 00:21:46] within their data set there were several hundred additional cases that were identified using this machine learning approach compared with what would have been predicted just using the standard algorithm. So I think it's quite exciting, it shows how using sort of a more unbiased approach, but still using variables that are generally present in the electronic health record can actually improve risk prediction. And this sort of approach might help us to do a better job of identifying you know, the really quite large number of people who do go on to develop incident CVD or have MIs without having the standard risk factors that we know about.

                                           Then I actually saw a second paper. So, for people who are interested in this sort of approach, there was a really nice review article that was published recently in JACC and this came out on 30th of May of this year, 2017, so just recently. The first and last authors of that paper are Chayakrit Krittanawong and Takeshi Kitai. The title of their paper is "Artificial Intelligence in Precision Cardiovascular Medicine." So in this review article, they discussed the potential of artificial intelligence to improve cardiovascular clinical care and they highlight both the challenges and the potentials. Overall, they emphasized how important it is for physicians to try to understand these new computational approaches. I think both so that we can harness the potential of these approaches, but also so that you know we who are in charge of patient care can understand the inherent limitations of these approaches.

                                           So, the overall message I think from both of these papers is that machine learning, it's really exciting, it has a huge amount of potential, but you know robots aren't going to replace physicians any time soon, so we really need to have physicians working in tandem with these sort of computational approaches to really harness their potential.

Naveen Pereira:              So Jane, that is fascinating and it's going to be especially important in the era of big data, where all medical centers eventually transitioning to electronic health records. So we have this wealth of information in the electronic health records, and we should do what large corporations have been doing, that is trying to individualize patient care by incorporating multiple parameters from the electronic health record to understand these patients better and come up with risk scores. About two years ago, we had published in the journal Studies in Health Technology and Informatics in 2015, a similar analysis trying to discern better incorporating multiple co-morbidities from the electronic medical record using machine learning techniques. We could improve predicting prognosis in heart failure patients and we found an 11% improvement in the area under the curve by using electronic health record data and incorporating co-morbidities by using machine learning techniques. So I think there's great promise for the future in medicine for this. 

Jane:                                  Yeah absolutely, and as you point out as more and more places are moving towards fully electronic health records, it's something that's actually relatively easy and very cost effective to implement, so it's definitely an exciting approach.

Naveen Pereira:              So Jane, this is very interesting talking to you about these various topics, but if I didn't pay particular attention to the author of the publication how can I access these manuscripts that we discussed?

Jane:                                  So actually all of the links and links to the full article on the PubMed abstract for all of the papers that we've talked about are on the website. So the podcast website you can access that at and if you go there you'll see a post for every episode of the podcast that we've done. So you can click on the episode you're interested in and then you'll find links to all of the papers and topics that we've discussed.

Naveen Pereira:              Wonderful Jane. Look forward to talking to you again next month.

Jane:                                  Me too. Okay, thanks Naveen.

Naveen Pereira:              Bye Jane.

Jane:                                  Bye.