PhD Candidate in Computational Neuroscience, Universitätsklinikum Hamburg Eppendorf
Anastasia Brovkin (AB): Thank you for taking the time! To start things off, how did you end up finding your research niche? Were you always interested in the questions you work on currently? Was it a linear path?
Andrew Zalesky (AZ): My PhD is in electrical engineering, and it was very much by chance that I ended up working in neuroscience. Specifically, my PhD focussed on establishing network models for an Internet technology called optical-burst switching. While this is very different from my current research, network modelling techniques remain universal to a certain extent regardless of whether they are used to model nervous systems or the Internet.
In 2007, near the end of my PhD, I took up a part-time position at the Melbourne Neuropsychiatry Centre to assist with analysing structural brain connectivity in patients with schizophrenia. This work sparked my interest in network neuroscience, and I was fortunate enough to be offered a competitive research fellowship from the Australian Research Council in 2008 to continue this research on a full-time basis. Throughout my PhD, I envisaged that I would work in the telecommunications industry, and so the move to neuroscience was drastic at the time—and certainly unplanned—but I remain extraordinarily grateful for the opportunity to change fields.
AB: Your bio mentions that you “adopt” an engineering mindset to study the brain. I have found that interdisciplinary work comes with its challenges (i.e., communication across fields with different scientific and technical terminology). Do you find it rewarding? Challenging? What contribution are you proudest of that directly resulted from bridging disciplines?
AZ: Yes, interdisciplinary research can be challenging at times. Engineering and Medicine research cultures are distinct in many ways. In the beginning, I found that adjusting to the quirks of a new field was frustrating at times. Opinions on which aspects of a research project are most valuable and important can sometimes diverge. Engineers emphasize quantitative rigor, but I’ve found that they can have difficulty grasping the broader dimensions—and biological practicalities—of the problem that they are trying to solve. The opposite can be the case for clinicians and medical researchers. Our lab is quite interdisciplinary, and I have accumulated lots of experience in helping cross-disciplinary teams to strike a balance between these positions.
I am proud of our recent work on personalizing brain stimulation therapy as an example of bridging disciplines. We developed a novel methodology to personalize a patient’s stimulation target using their fMRI scans. In December 2021, our technology was approved by Australia’s Therapeutics Good Administration and licensed to Australia’s first clinic to offer personalized TMS therapy. It has been used to guide the treatment of 17 patients to date.
AB: What was the best advice you ever received from a mentor? What is your best advice you can give to early career researchers (ECR) that want to succeed in your field?
AZ: I get the sense that more and more is now expected of ECRs in terms of engagement, teaching, leadership and service. I think it is important for ECRs to focus on the basics of research and building a track record in research excellence. Persistence is important: Don’t be disheartened by a paper rejection or unsuccessful grant application. It took me three attempts to get my first major grant.
It has been a difficult couple of years for ECRs. I think that COVID-related restrictions and the lack of networking opportunities have had a particularly detrimental impact on ECRs. It’s fantastic, then, that so many ECRs have finally been able to experience OHBM in-person for the first time this year!
AB: You co-authored Fundamentals of Brain Network Analysis with Prof. Alex Fornito and Prof. Ed Bullmore, which serves as a wonderful primer into network neuroscience, and I’ve seen a lot of folks starting out in the field use it. What was your personal motivation for taking part in this project? How did it come about? Given the methodological progress in the field, is another book on the horizon?
AZ: Thanks for the positive feedback about our book, Anastasia. Alex and Ed are longstanding collaborators. Alex and I also used to be based in the same research centre. Alex came up with the idea of writing the book, and I was very glad to contribute. Alex is a fantastic collaborator. One of our motivations was to introduce researchers and grad students to the rapidly developing field of brain connectomics. There were no introductory texts at the time, and newcomers had to trawl through a maze of papers to get up to speed. It is great to see that our book is now widely used and forms the basis of graduate courses. We’ve been thinking about a second edition, but there are no concrete plans yet. Writing a book is a huge undertaking. Hopefully the second edition will be easier.
AB: Your most cited paper is on network-based statistics. As it was written 12 years ago, what would you say have been the key methodological advances in this field since then?
AZ: Twelve years is a long time in neuroimaging. There have been many important advances and conceptual shifts over the last decade in neuroimaging methodology and statistics. Too many to mention here! A shift from statistical inference to predictive modelling is perhaps the most salient in my opinion. And a renewed emphasis on the importance of effect and sample sizes as well as adequately powered studies also emerged in the last decade.
I sense that current methods might only be detecting the tip of the iceberg and that effects are in fact more widely distributed throughout the brain than we think. Stephanie Noble has some recent work on this topic. With larger sample sizes and the ability to detect vanishingly small effect sizes, it will be important to decide when an effect is small enough to be ignored, even though it may be statistically significant.
AB: When it comes to applied methods, do you have any pet peeves, i.e., do you see a lot of common method fallacies researchers in the community make? What is the most frequent one? How can they be avoided?
AZ: I don’t like the term “fallacy” in this respect. When dealing with high dimensional and noisy data, it is difficult to be certain that one method is universally better than its counterparts. Be wary of strongly opinionated methodologists who are fervent about telling you which method you must use to analyse your data. For example, while the diffusion tensor is a dated fiber orientation model, using the model is not a fallacy. Similarly, I think that deterministic white matter tractography still has a role to play in mapping connectomes, despite its criticisms. Pet peeves? The ongoing squabbles between MATLAB and Python users that flare up on Twitter and elsewhere are kind of pointless in my opinion. Surely access to adequate computing resources, particularly in disadvantaged regions, and researcher training are more important issues.
AB: Given the relevance of your research to the field of psychiatry, do you often see any successful translation in the clinical setting? What would you say is the biggest barrier or limitation between research and clinical application? What would you say was the biggest success over the past decade?
AZ: We recently developed a methodology to personalize an individual’s brain stimulation coordinates based on their fMRI scans. In December 2021, our technology was approved by Australia’s Therapeutics Good Administration and licensed to Australia’s first clinic to offer personalized TMS therapy. It has now been used to guide the treatment of 17 patients to date. Clinical translation is challenging and opening the clinic would not have been possible without the dedication and hard work of Luca Cocchi and Bjorn Burgher. I am proud of the clinic because it is an example of successful translation in a clinical setting of a basic fMRI methodology. Neuroimaging data is noisy and many current image analysis platforms in the field are not sufficiently robust and generalizable to the kinds of data that are acquired clinically.
AB: Your lab released the Melbourne Subcortex Atlas. What was the motivation behind this? Would you argue a whole-brain connectome approach is preferable to other methods? What are some limitations? What are the benefits?
AZ: Strangely enough the majority of modern brain parcellation atlases do not include subcortical territories. We noticed that many researchers would simply not include subcortical nodes when mapping connectomes. But the subcortex is such as crucial brain structure, both in health and understanding brain disease, and ignoring it when establishing a whole-brain, systems representation of brain network organization seemed crazy to us. Ye Tian has now incorporated the Melbourne Subcortex Atlas into numerous cortex-only parcellation atlases to enable mapping of truly whole-brain connectomes. Well, except for the cerebellum and brain stem – we still need a functional parcellation for these structures. I don’t think that the whole-brain connectome approach is necessarily preferable compared to other methods. It really depends on your aims and the questions that you are aiming to address. If you are specifically interested in studying fronto-striatal circuitry, for example, a whole-brain approach might not be warranted.
AB: Would you say open science and reproducibility play a big role in your lab? What would you like to see more of in the Network Neuroscience community with regards to code sharing and methods documentation? What about the Neuroimaging community overall?
AZ: I think that the field has come a long way in the last decade with respect to open science and reproducibility. Code and data sharing is now the norm, whereas when I started in the field, the norm was “code available on reasonable request”. We need to thank Russ Poldrack and colleagues for their pioneering efforts. About a decade ago, I remember being told by a senior colleague that our data is like moon dust and it needs to be protected from other researchers. Times have changed. Regarding reproducibility, I think we need to remember that a single failed replication study does not necessarily overturn the original finding. I’m a fan of equivalence testing and I think that its more widespread use in neuroimaging could potentially improve reproducibility.
AB: What is the take-home message of your keynote at OHBM this year? What are the most exciting avenues of research you’ve explored recently?
AZ: I’ll be presenting some of our new research on characterizing brain-body interactions across the lifespan. As neuroimagers and neuroscientists, we often overlook the fact that the brain is interconnected to virtually every other organ and body system. I hope I can convince OHBMers of the importance of understanding how the brain influences, and is influenced by other body systems. This is needed if we are to establish a systems-level characterization of brain function and structure.
AB: That sounds exciting! Finally, what do you enjoy outside of research?
AZ: Thanks Anastasia. I spend way too much time in front of a computer screen. Anything outdoors in my spare time.
AB: Thank you for your time and for the interview!