Multi-Omics: Cathepsin C Flags Cancer Therapy Heart Risk
Troponin and NT-proBNP are the blood tests oncologists order when they worry about the heart during anthracycline or trastuzumab treatment. Neither predicts well who will lose pump function. A University of Pennsylvania team, working with Pfizer, tested whether a multi-omics approach could do better in 547 patients with breast cancer. They paired plasma protein profiling of nearly 3,000 targets with rapid LC-MS metabolomics at three timepoints and matched every draw to core-lab echocardiography.
Key Takeaways
Multi-omics plasma profiling in 547 breast cancer patients found candidate cardiotoxicity markers that carried more risk signal than troponin or NT-proBNP.
Higher baseline plasma cathepsin C was linked to lower risk of cancer therapy-related cardiac dysfunction, but the finding still needs an independent cohort.
Serial blood draws matched to same-day cardiac imaging let drug developers separate diagnostic biomarkers from predictive ones in cardiac safety studies.
A Longitudinal Multi-Omics Design in Cardio-Oncology
About 61% of the Penn cohort received doxorubicin-based chemotherapy; the rest got trastuzumab with or without an anthracycline. Proteins were read on the Olink Explore 3072 proximity extension assay, which reports relative NPX units, and Sapient Bioanalytics ran the metabolome. The statistical design is worth copying. Mixed-effect models tested each analyte against eight echocardiographic measures, including left ventricular ejection fraction (LVEF) and longitudinal strain, at the same visit (diagnostic) and the next one (predictive). Within a median of 7 months, 14% of patients developed cardiac dysfunction.
Key Findings
203 proteins and 16 metabolites tied to cardiac measures: All passed false discovery rate (FDR) correction, and proteins rising with systolic function clustered in deubiquitination and catabolic pathways.
Cathepsin C (CTSC) stood out: It tracked LVEF (FDR P = .017), strain, and left atrial volume, and each baseline NPX unit carried a hazard ratio of 0.61 (95% CI 0.41 to 0.90) for incident dysfunction.
Predictive signals concentrated in strain: In lagged models, 113 proteins forecast longitudinal strain at the next visit; none forecast LVEF.
Where the metabolome pointed: Aspartic acid, N-acetylglutamine, acetylasparagine, alanyl-alanine, and prolyl-glycine tracked both LVEF and strain (FDR P < .001).
Figure 1. LOESS curves with 95% confidence bands tracing echocardiographic measures over two years of breast cancer therapy. Red lines mark participants who developed cardiac dysfunction; blue lines mark those who did not. Panels span systolic, structural, and diastolic measures, from LVEF and strain to left atrial volume index and Ea/Ees coupling. Adapted from Ky et al. (2026), European Heart Journal.
Cathepsin C and the Proteomic Signal
Cathepsin C, also called dipeptidyl peptidase 1 (DPP1), is a lysosomal protease that activates neutrophil serine proteases. Prior studies tied it to unstable plaque and ischemia-reperfusion injury, so the protective direction seen here is counterintuitive, and it merits a closer look now that a DPP1 inhibitor is approved for bronchiectasis. Adding CTSC to a clinical model only nudged the concordance index from 0.66 to 0.68. With discovery and testing done in one cohort, treat the hazard ratio as a hypothesis until it replicates.
Amino Acids, Not Acylcarnitines
The metabolomic result cuts against expectations. Targeted work from Asnani and colleagues (2020) in 38 treated women pointed to citric acid cycle intermediates, yet acylcarnitines and TCA intermediates on this panel fell short of significance. Aspartic acid instead tracked LVEF and strain, then LV mass index at the next visit, consistent with animal data linking aspartate to cardiomyocyte hypertrophy. The cross-omics analysis was the paper's thinnest part: protein-metabolite correlations peaked at about 0.4 in absolute value. A comparable plasma design recently flagged VDAC1 as a candidate marker in atrial fibrillation.
Why Same-Sample Integration Changes the Read
Correlations that weak aren't surprising when two layers come from separate aliquots on unrelated chemistries, since pre-analytical and batch noise then enter every cross-layer comparison. Running MS-based proteomics and metabolomics from one plasma extract, which the Omni-MS workflow at Dalton does in a single injection, strips out that noise so a modest correlation can be read as biology. For serial-sampling safety cohorts where banked plasma is finite, single-injection multi-omics profiling is the more practical route into multi-omics biomarker discovery.
Frequently Asked Questions
What is multi-omics used for in cardio-oncology?
Multi-omics measures two or more molecular layers, usually proteins and metabolites from blood, to find markers that a single assay would miss. In cardio-oncology it helps flag patients at risk of heart damage from cancer drugs and points toward cardioprotective targets.
Can a blood test predict heart damage from anthracycline or trastuzumab treatment?
Not reliably yet. Troponin and NT-proBNP failed to predict cardiac dysfunction in this cohort. Baseline cathepsin C did, but its added value was small and it has not been tested in an independent cohort.
How many samples does a multi-omics biomarker study need?
Event count matters more than enrollment. This study sampled 547 patients up to three times, yet only about 14% developed dysfunction, which limited statistical power. Budget for serial draws matched to imaging plus a separate validation cohort.
Conclusion
It's now reasonable to say that circulating proteins and amino acid metabolites move with cardiac function during anthracycline and trastuzumab therapy. Whether cathepsin C can stratify patients isn't settled. For oncology programs with a cardiac safety question, the practical move is to bank serial plasma alongside serial imaging before the first dose.
Related Reading
See how we run these analyses in one lab: Dalton's multi-omics CRO services.
Citation
Ky, B., Xia, C., Ko, K., Hyde, C., Smith, A. M., Rhee, J.-W., Tow-Keogh, C., Tierney, C., Long, T., Zhang, L., Liu, P. P., Wilcox, N. S., Dang, V., Armenian, S. H., Jain, M., Mangipudy, R., & Vaidya, V. S. (2026). Cardiac remodelling and dysfunction in cancer patients receiving cardiotoxic therapies: Proteomic and metabolomic profiling. European Heart Journal, 47(35), 5010-5025. https://doi.org/10.1093/eurheartj/ehag487
Note
This blog post summarizes findings from the above-cited research. Figures are adapted from the original publication. For full details, please refer to the source article.
By Seungjun Yeo, CEO at Dalton Bioanalytics. Specializing in multi-omics mass spectrometry for drug discovery and biomarker research.