The following is the first entry in a new Q&A series highlighting selected Clinical Proteomic Tumor Analysis Consortium (CPTAC) researchers and their work. Join us as we discuss molecular and computational techniques in biology with Natalie Clark, PhD from the Proteomics Platform at the Broad Institute. Transcript is edited for clarity.
Q: Can you describe your experience integrating molecular and computational techniques in biology?
NC: I started on this hybrid experimental-computational journey during my PhD at North Carolina State University, where I worked in plant biology. My research utilized a plant called Arabidopsis and during that time I was not only working in the lab … I was also doing data analysis …and I found that I really like to be able to do both. I like to be able to be in the wet lab, do the experiment, take the data, and then do the analysis all by myself--that was really fulfilling for me. I continued that (hybrid role) during my post-doc at Iowa State University (ISU), where I started to do proteomics, and now at the Broad Institute I’ve transitioned into a fully computational role. So, while I don't do any of the experiments anymore, I still have that foundation and I find that it helps me when I do the analysis, or as I develop these analysis methods since I have an idea of the biology and the techniques.
Q: Please elaborate on your role in the CPTAC-LUAD project.
NC: The CPTAC-LUAD (Lung Adenocarcinoma) project at the Broad Institute 
involves integrating various genomics and proteomics data sets from different cohorts. This project includes a discovery cohort (Cell, 2020) and a confirmatory cohort… I [joined] after the first cohort had been published (graphical abstract pictured right) and the second cohort was being finalized. My first project was to harmonize these two cohorts because while they were processed and collected in similar ways there was a gap in time… so additional processing was required to ensure they were comparable. Moreover, we're collaborating with the ICPC-Taiwan team and, similarly, [they also have two cohorts]. If you put these four cohorts together, we have over 400 patient samples, or over 800 unique samples when you separate the tumors and the [normal adjacent tissue samples], which is definitely the largest data set that I've ever worked with personally. Integrating everything across different cohorts, across different countries, and across different omes has been challenging but fulfilling and I think at the end of the day, this is going to be a really exciting and intriguing collaboration.
Q: Can you mention aspects of the development and implementation of any analysis methods you're applying at present?
NC: We employ a number of different techniques in our group, and I would say the main goal is to integrate all of these different omics data together. The main technique we use is a suite of tools called PANOPLY developed by Dr. D.R. Mani at the Broad Institute. PANOPLY allows us to normalize, filter, and combine data via methods such as correlation analysis, association analysis, clustering, and outlier analysis (overview pictured below). Additionally, we're working on incorporating new techniques into PANOPLY, such as network inference methods like SC-ION, a method I developed during my post-doc, which infers regulatory networks based on multiple omics types together.
The development of SC-ION was driven by the need to process large-scale data like you find in transcriptomics and proteomics and use those data to infer unsupervised networks. In essence, you supply [SC-ION] with data, like gene expression or protein expression data, and based on a mathematical model, you can predict if one gene regulates another, potentially with activation or repression. [SC-ION] basically allows you to supply two omes-- one is the regulatory one and the other is the target. It allows you to create networks where you have protein-predicting transcripts or phosphoprotein-predicting proteins, in pretty much any combination you can think of. Additionally, SC-ION provides the network data in a format suitable for visualization in applications like Cytoscape or for further analysis to identify key regulators in the network.
Q: Are there any unmet needs or challenges in your field that you hope to resolve during your career or that you foresee for the next generation of young scientists?
NC: One key area is the integration of different omics data. We've made significant progress in combining multiple omics datasets, such as through tools like PANOPLY and SC-ION. However, there are still many methods that claim to be multi-omics that only analyze each omics dataset separately. What we need are more methods that truly integrate multiple omics datasets simultaneously and harmonize them. Another place where I think that we need to innovate is in the single-cell landscape, specifically the single-cell proteomics landscape. It's a challenging area because proteomics data differs in distribution and characteristics compared to genomics data, and there's also a need for optimization on the experimental side.
Q: What is an example of a project you worked on that inspired you and/or changed how you approach your work in computational science?
NC: One of the ISU Committee for the Advancement of Women and Gender Equity’s key initiatives is to produce a report every five years focused on enrollment, retention, salaries, and experiences across different groups of community members, with the goal of identifying inequities. Before I joined, these reports were just white papers in PDF format, posted on the website, and rarely seen by the community. The committee aimed to change that by creating a user-friendly website where the report could be displayed for everyone to see and continuously updated. I was one of the main people on the committee who had data analysis experience; I collected and analyzed the data and generated visualizations (pictured below) to make the data more accessible and understandable to a broader audience.
The committee uses this report to bring up issues found in the data. For example, public universities like ISU are required to use census classifications for things like race and gender. From a race standpoint… if you're a person that identifies as more than one race, they [assign] you to a category called “multiple races.” From a gender standpoint [the university] generates no data on gender nonconforming, nonbinary, and transgender students. And so, when you're trying to look at the experiences of your community members, it's difficult because you're already not accounting for some of your most unique members. My experience with the Committee for the Advancement of Women and Gender Equity has definitely influenced how I approach projects like CPTAC-LUAD and others that involve clinical data and serves as a reminder to approach data analysis with ethical considerations.
Q: If a young and aspiring scientist were looking to follow a similar path to yours and reach the point you're at in your career, what advice would you offer them?
NC: My main advice would be to follow the science that interests you. I don't want to advocate and say you need to do the hybrid to do what I do, or that you need to be in the lab and be computational to end up where I am… I think a lot of times people feel pressured that they need to do both. I think that you should do what makes you feel most fulfilled.
That said, no matter where you go in life, I definitely recommend learning a programming language because computer-based analysis is fundamental in today's scientific landscape… and a lot of great resources are available online for free. If possible, I also think a foundation in statistics is really good no matter what you do, whether you're in the lab or doing computational work.
*Transcript edited for clarity*