A blood-based classifier for differential diagnosis of feline IBD and enteropathy-associated T-cell lymphoma

Josephine Tsang^1†, Surya Vishnubhatt^1†, George Courcoubetis¹, Karim Mrouj¹, Sushmita Sen¹, Chewon Yim², Hyoju Yi², Beeham Lee², Sheena Kapoor¹, Jerry Cromarty¹, Sungwon Lim^1,2, Jamin Koo^1,2,3,*, Ilona Holcomb¹*

¹ImpriMed, Inc., 3980 Fabian Way, Palo Alto, CA 94303, USA

²ImpriMedKorea, Inc., Seoul 03920, Republic of Korea

³Department of Chemical Engineering, Hongik University, Seoul 04066, Republic of Korea

† Author equal contribution

Introduction

Differentiating feline IBD from enteropathy-associated T-cell lymphoma (EATL) remains a clinical challenge owing to overlapping presentations and the stress and cost of invasive sampling. Building on prior identification of disease-specific transcriptomic and methylation signatures in minimally invasive specimens, we expanded the cohort and developed a multi-omic blood-based classifier.

‍

Methods

The cohort consisted of 60 cats (30 EATL, 30 IBD; histopathologically confirmed) with paired RNA-seq and enzymatic methyl-sequencing (EM-seq) profiles from whole blood, together with clinical and demographic data. Candidate biomarkers were identified in a 40-sample training set using DESeq2 (DEGs) and DSS (DMLs), refined through elastic-net stability selection over stratified resamples, and curated by biological review. A separate clinical-demographic model was developed to provide a baseline risk score. A locked panel of 10 genes and 10 methylation features, together with three pre-specified classifiers, was evaluated on the remaining 20 samples. Performance was assessed using sensitivity, specificity, and accuracy with 95% confidence intervals.

‍

Results

Early leading classifiers met pre-specified thresholds of ≥70% sensitivity, ≥75% specificity, and ≥70% accuracy on the held-out test set. The majority of curated biomarkers replicated in direction and magnitude between the training and held-out sets.

‍

Conclusion

Initial evaluation supports the diagnostic utility of a blood-based classifier for distinguishing feline IBD from EATL. Ongoing work focuses on improving performance, validation, and development of a clinically deployable diagnostic assay.