Masterclass in advancing precision medicine

Crohn’s disease is a chronic, autoimmune inflammatory bowel condition with no available cure, that can present in pediatric patients who may later develop serious complications such as fistulas and fibrostenosis. Prognostic biomarkers to predict such complications would be valuable tools to optimize care for pediatric patients, but are currently lacking. This webinar will describe how a combination of analytically validated, high-throughput proteomics and transcriptomics can be used to identify prognostic markers with a high degree of accuracy in predicting pediatric patients who will develop complications.

Join this free on-demand webinar with Dr. Andrés Hurtado-Lorenzo, VP of Translational Research at the Crohn's and Colitis Foundation, as he discusses how high-quality protein biomarker discovery provides deeper insights into disease biology and actionable data for improved care.

Read on for highlights from the Q&A discussion and register now to watch the webinar on-demand.

Did you evaluate whether integrating clinical data with the prognostic biomarker panels can influence the performance metrics?

AHL: The initial academic consortium did integrate clinical data and demographic data with the transcriptomics data and that led to the composite risk model that basically we used as an example. Then when we started our own efforts with machine learning, we did that. We integrated the clinical data together with transcriptomics and even microbiomics. However, we didn't see any improvement in our risk model. We're going to see whether we can improve performance with the integration of clinical data.

Did you investigate whether the biomarker signatures are stable over time?

AHL: Yes, another great question. So, we are now in the process of answering that question, we have just received approval to access the longitudinal samples that are stored for the risk study in the IBD Plexus platform. And we're going to start the study. However, I can give you a spoiler which is that we were able to analyze some samples of patients that were recruited at the time of diagnosis. When they enrolled, they already had complications, and we were able to overcome them using our proteomics panel. We were able to stratify the population in the different phenotypic groups, the B2, the B3, and the B2/B3 using the prognostic panel.

So what that suggests is that perhaps the signatures are longitudinally stable since we have already found them in the patients who already have complications.

Can your prognostic biomarkers predict the time to event?

AHL: We did group the patients according to the time when they developed the complications. So we divided the groups into early, which is where patients were diagnosed with the complications within the first year, then mid, when patients were diagnosed with the complication in the second year, and late, when the patients were diagnosed with the complication after the second year.

We run the machine learning algorithm, this is with our proteomics panel, and our preliminary data indicate that actually, the performance is much better for early and mid, basically within the first to second year. This is where the performance and the prediction seem to be much better with one of the machine learning algorithms that we used. And we used another machine learning that actually helped us predict the late events, but the best performance seems to be the first two years, according to these initial studies.

How does the panel compare with fecal calprotectin in terms of performance?

AHL: I think we have to define the different types of contexts of the use of biomarkers. So, in this case, fecal calprotectin basically is a diagnostic and eventually a mucosal healing and monitoring biomarker. So as far as my knowledge, fecal calprotectin is not a prognostic biomarker, so it would be difficult to know the performance measurements of fecal calprotectin as a prognostic because I haven't seen that data. It is rather a diagnostic biomarker of mucosal healing, but here we're talking about complications, fibrosis, and fistulas.

What machine learning algorithms did you use with the Olink discovery data?

AHL: I really would like to apologize to the audience given the fact that we are submitting this for intellectual property. So, I was asked by our IP lawyers not to disclose the machine learning classifiers, nor the identity of the analytes that are the biomarkers that we are talking about, unfortunately. So, I really apologize, but once the patent is published, we will be publishing this data and making it more widely available.

To learn more about microspectroscopy, watch the full webinar here>>

SelectScience runs 10+ webinars a month across various scientific topics, discover more of our upcoming webinars>>