Next Gen in 10: How the Vaginal Microbiome Could Revolutionize Cancer Detection

In this episode of Next Gen in 10🎙️, Sabrina chats with Dollina Dodani, a PhD candidate in Bioinformatics at UBC. Under the supervision of Dr. Aline Talhouk, Dollina is exploring how the vaginal microbiome 🧫 could be used to improve early screening and detection of endometrial cancer. She shares how she’s using machine learning and transfer learning to build a “General Vaginal Microbiome Classifier” — a computational tool that could transform how we diagnose gynecological diseases. From decoding complex data 📊 to uncovering microbial clues 🔍, this episode dives into the intersection of biology, AI, and women’s health.

For more information on the Gynecologic Cancer Initiative, please visit https://gynecancerinitiative.ca/ or email us at info@gynecancerinitiative.ca

Where to learn more about us:
Twitter – @GCI_Cluster
Instagram – @gynecancerinitiative
Facebook – facebook.com/gynecancerinitiative
TikTok – @gci_gosh

Transcript

00:00:02 Intro 

Thanks for listening to the GOSH podcast—The Gynecologic Oncology Sharing Hub. We share real, evidence-based discussions on gynecologic cancers, featuring stories from patients, survivors, researchers, and clinicians. Our podcast is produced and recorded on traditional unceded territories of the Musqueam, Squamish, and Tsleil-Waututh Nations. It is produced by the Gynecologic Cancer Initiative, a BC-wide effort to advance research and care for gynecologic cancers.  

00:00:35 Sabrina 

Hi everyone! My name is Sabrina, and I would like to welcome you back to our new segment on the GOSH Podcast called Next Gen in 10 where we feature GCI trainee research. Today we are joined by Dollina Dodani who is a PhD candidate in bioinformatics at UBC. Her research interests lie in developing computational techniques to analyze high-dimensional biological data and identify biomarkers that tailor patient care. Under the supervision of Dr. Aline Talhouk, Dollina’s research explores the potential utility of the vaginal microbiome to improve screening and detection of endometrial cancer. Welcome Dollina! 

00:01:16 Dollina 

Thank you. Thank you for having me. 

00:01:18 Sabrina 

I was thinking we could start sort of to give our audience a background of what your research is. So would you mind telling us a bit about the background of your research topic and or the gap that your research is aiming to fill? 

00:01:31 Dollina 

Yeah, of course. So as part of my research, we're mostly focusing on endometrial cancer or cancer of the uterus, just the cancer that starts in the inner lining of the uterus. And it is the most common gynecological cancer in most of the developed countries. And incidence is rising across the world as well. But unfortunately, there are no ways to screen for this cancer. So a woman or an individual with a uterus will present symptoms, and these symptoms are quite nonspecific. It's abnormal uterine bleeding, pelvic pain. And unfortunately, they are required to undergo a biopsy, which is an invasive, very painful procedure. And most of these women, less than 5% of these women actually end up having the cancer that they're looking for. So you can imagine how overly used a biopsy is and how many women have to go through it when they don't really have to go through it. So my research really focuses on trying to identify biomarkers from other data sources that can flag the need for a biopsy and can flag high risk women who need a biopsy. So that's what we're working on and we're using vaginal microbiome to do so and other risk factors as well. 

00:02:48 Sabrina 

Interesting. Very interesting. So what's your research question or the objectives? 

00:02:56 Dollina 

The objective is, is the vaginal microbiome obtained from a minimally invasive tool like a swab, a vaginal swab? Can it be reliably used to identify women who need a biopsy and who may have cancer? That is the overarching research question, and we're using machine learning and bioinformatics to do this. 

00:03:18 Sabrina 

Okay, so I think you mentioned a little bit there what your methods are. Could you elaborate a bit on what the methods are or how exactly you're going about answering your research question? 

00:03:29 Dollina 

Yeah, so we have, a lot of studies have studied the vaginal microbiome in the context of endometrial cancer across the globe. And they all come up with their own data set, and they all process it, and they come up with a signature. So what we're trying to do is leverage this wealth of publicly available data that we have from across the globe, from across different geographic locations, ethnicities. And we're using machine learning to integrate all the data, to harmonize the data, and come up with a predictive model for incoming new samples. So it's a data that our lab is collecting as well. We want to validate those machine learning predictive models on new data. 

00:04:10 Sabrina 

That's great. Would you mind elaborating on sort of what machine learning is for anyone listening who doesn't understand and how that can improve your disease prediction? 

00:04:21 Dollina 

Yeah, machine learning is basically fundamentally a statistical technique where you train a model to learn from existing data. So you're forcing this tool to look at various examples of what cancer looks like, of what benign conditions look like, what healthy aging looks like. And then you take in a new data point and you see where it fits in, in the pattern that the tool has learned with all these several existing data. So that's what machine learning is, you're trying to learn from the data that you currently have access to. And you learn patterns and every new incoming data point or incoming patient, you see where it fits in this pattern to predict whether they would be at high risk for endometrial cancer or they're just benign. It's normal aging. They probably have a benign condition. You kind of learn to predict that with a probability and a confidence as well. 

00:05:20 Sabrina 

Interesting. Yeah, I think that's a great definition. I feel like machine learning is talked about all the time, but people don't necessarily understand what it means. So that'll be really helpful for our listeners. Would you mind sharing a bit about what you've found so far, like what your results are and anything significant? 

00:05:38 Dollina 

Yeah, so we integrated data from national microbiome data from five existing studies that have published data on this. And what we found is that there is a very strong signature in the microbiome that can help identify those who are benign or healthy versus those who need a biopsy and could be at high risk for cancer. And these come down to very specific strains of the bacteria that live in a vagina and uteruses and things like that. So one particular strain that is very indicative of how healthy you are is lactobacillus. It's very commonly found in the probiotics that we take today that are commonly found on drugstore shelves. That is a very strong indicator of how healthy your body is. And in terms of those who were with cancer, there is a depletion of this particular strain. And the other more harmful commensal bacteria start to pop up in those cases. So we're trying to validate that model currently on new data sets that our lab has successfully set up through our clinical trials. Yeah, so we're going to see how well those models validate there. 

00:06:55 Sabrina 

Interesting. That's super interesting. Would you mind telling us a bit about why you went into this field and why you were interested in focusing on the vaginal microbiome in connection to diseases like endometrial cancer? 

00:07:11 Dollina 

I think when I first started my PhD, the microbiome was such an up and coming buzzword. It was everywhere. But very limited work was being done in the gyne space. Sadly, actually, most of it is done in colorectal cancer and in other spaces. So I think it was very exciting to be part of such pioneering work that Dr. Aline Talhouk is involved in and is trying to bring to the clinic to improve women's health. And also importantly, there was very limited work done in machine learning to bring machine learning to the space as well to make this more clinically accessible. So these are the things that really got me excited. And like I mentioned, our lab has an amazing clinical trial going on to recruit women with and without abnormal uterine bleeding for the same purpose. So it was very exciting to be part of this work and to bring machine learning to the space. 

00:08:09 Sabrina 

Fantastic. I think what you mentioned about the microbiome being a buzzword is absolutely true. And I think it's even cooler that you're not just doing the general gut microbiome that most people look at. You're looking at the vaginal microbiome, which is even more niche, but even, I think, very cool. 

00:08:25 Dollina 

Yeah. Can you imagine for the longest time, the vaginal space was considered to be sterile. We didn't even know there was a microbiome there until like a decade ago. So I think we've come a long way since then. 

00:08:36 Sabrina 

Yeah, absolutely. So you've worked with the microbiome data for a little while now. Could you share some of the biggest challenges you've run into while working with this data? 

00:08:48 Dollina 

I think because it's so understudied, the tools that we're mostly using were developed for the gut microbiome investigation, and the assumptions that these tools use are more fitting to the gut microbiome. So I think that was the biggest challenge for me, was to really understand how different the vaginal microbiome is compared to the gut microbiome, because it is very, very different in composition and structure. Other than the computational tools, also like the awareness of the vaginal microbiome is quite limited. How do we process this? How do we collect this in a more reliable manner? Because you can't use the same protocols that you did for the gut microbiome, it is very different. So just understanding those kind of limitations and technicalities was quite a challenge. 

00:09:36 Sabrina 

Yeah, that makes sense. I think it's always hard when you're trying to apply techniques that were created for something else to something new. But kudos to you for exploring all of that and working very hard to get such important data. So you mentioned in some of your research that I've actually heard that you have this whole general vaginal microbiome classifier. How do you hope that this will change the way doctors diagnose or understand gynecologic diseases in the future? 

00:10:09 Dollina 

Yeah, so just a little bit background on the generalized vaginal microbiome classifier. What we're trying to do is that we're trying to leverage data from other disease contexts as well. So we're looking at data from breast cancer, ovarian cancer, endometriosis, PCOS, fibroids, and we're trying to learn trends in all of this data and see if we can apply it to endometrial cancer. Because if we know what everything else is, we can probably transfer that knowledge and bring it to endometrial cancer. And that's transfer learning. You can almost think of it as learning how to play the piano, and then you transfer that knowledge and learning how to play the guitar. And that's what we're trying to do with our models as well. So that is the, that we hope to call is the generalized vaginal microbiome classifier. And what we're hoping that would... The impact that we're hoping that would have is not only can you use it for endometrial cancer classification and detection, you can also reuse that model that we're hoping to develop at the uterine Health Research Lab for other diseases of interest, such as ovarian cancer and breast cancer, and see if it can be applied for the earlier detection of those conditions as well. So we're really hoping that this open source model would really change the way the vaginal microbiome is perceived in the gyne space and really have an impact on not only endometrial cancer but also benign conditions that are often misdiagnosed, like enemy choices or fertility issue. We're hoping that this would have an impact not only in endometrial cancer but also gyne conditions, other gyne conditions that are often misdiagnosed, like enemy choices. 

00:11:44 Sabrina 

So from my understanding, do you imagine that someone could come in that's having some sort of gynecologic issue, have their vagina swabbed, and then that would give indications as to where to go with disease? 

00:11:58 Dollina 

Yeah, so for now, the classifier is mostly just looking at endometrial cancer. That's what it is trained to do, and that's what it's designed to do. But because it is a transfer learning technique, and it's been trained on different data sources, from other contexts as well. You could potentially retrain it as well for other things. 

00:12:19 Sabrina 

Interesting. So like long, long term, there could be even more implications for the work you're doing. Fantastic. Okay, to wrap up this episode, we always ask our guests if they could say one thing to everyone who will listen to this podcast about your field of research, it can be anything. What would you want them to know? 

00:12:40 Dollina 

I think what I've realized through the course of my work with the Uterine Health Research Lab is that it is so important to discuss symptoms and discuss what you're going through, what your body is going through. And often like women are facing a late diagnosis because they weren't open with what they were going through. And also me, like I'm often quite conscious about discussing these things. But I think I would, I would like to ask for almost like a mindset shift and discussing gynae symptoms and talking about gynae cancers as well would be really, really cool to see the next generation doing. And also just participating in research. Like I was mentioning that our lab has such an amazing community advisory group that help us with research. So it would be great to just put that word out there and get more women and more advocates for this to offer their perspective as caregivers or survivors. I think it all makes a difference. 

00:13:41 Sabrina 

Fantastic. I think those are great, two great calls to action for our listeners. Thank you so much for joining us today and sharing all about your research and everything you've learned. Super interesting work. Thank you and I hope that you continue doing super amazing work. 

00:13:57 Dollina 

Thank you. Thank you for having me and setting up this great platform. 

00:14:02 Outro 

Thanks for joining us on the GOSH Podcast. To learn more about the Gynecologic Cancer Initiative and our podcast, make sure to check out our website at gynecancerinitiative.ca.