e21148 Background: Protein glycosylation is the most abundant and complex form of post-translational protein modification. Glycosylation profoundly affects protein structure, conformation, and function. The elucidation of the potential role of differential protein glycosylation as biomarkers has so far been limited by the technical complexity of generating and interpreting this information. We have recently established a novel, powerful platform that combines liquid chromatography/mass spectrometry with a proprietary artificial-intelligence-based data processing engine that allows, for the first time, highly scalable interrogation of the glycoproteome. Methods: Using this platform, we interrogated 694 glycopeptide (GP) and non-glycosylated peptide transitions derived from 74 serum proteins in pre-treatment peripheral blood samples from a cohort of 316 individuals with non-small-cell lung cancer (NSCLC) (128 females, 187 males, 1 with unknown sex, median age 66 years, age range 31-89 years, stage 0-4 N’s: 1 / 99 / 80 / 84 / 49, 3 missing) and a comparison cohort of 194 healthy control samples (102 females, 92 males, median age 52 years, age range 30-63 years). Age- and sex-adjusted differential expression analysis for 596 normalized biomarkers were performed to evaluate statistically significant differential abundances using an FDR-adjusted q-value of 0.05 as a cutoff. Repeated five-fold cross-validated LASSO-regularized logistic regression was performed to create a multivariable classifier that predicts whether a serum sample belongs to the healthy or NSCLC cohort. Results: We identified 432 biomarkers with significant abundance differences at FDR ≤ 0.05 between samples with NSCLC and healthy controls. Using 70% of the complete cohort (balanced by case/control membership, NSCLC stage, sex, and age quartile) as a training set, we selected a total of 375 glycopeptide and non-glycosylated peptide biomarker features that remained differentially expressed at FDR-adjusted q-value ≤ 0.05 as input into a LASSO-regularized multivariable classifier. This resulting in a 19-biomarker model exhibiting an accuracy of 94.8% (96.9% sensitivity, 91.2% specificity) and AUC of 0.989. This classifier was validated in an independent test set comprising the remaining 30% of subjects, yielding an accuracy of 94.5% (95.5% sensitivity, 93.0% specificity) and AUC of 0.975. Sensitivity in the test set was 100% / 96% / 99% / 96% / 94% / 10%, in stages 0-4 and missing, respectively. Conclusions: Our results indicate that glycoproteomic biomarkers can be leveraged as a strong liquid biopsy-based screening tool for patients at high risk of NSCLC, as an alternative to imaging modalities.

The clinical success of immune-checkpoint inhibitors (ICI) in both resected and metastatic melanoma has confirmed the validity of therapeutic strategies that boost the immune system to counteract cancer. However, half of patients with metastatic disease treated with even the most aggressive regimen do not derive durable clinical benefit. Thus, there is a critical need for predictive biomarkers that can identify individuals who are unlikely to benefit with high accuracy so that these patients may be spared the toxicity of treatment without the likely benefit of response. Ideally, such an assay would have a fast turnaround time and minimal invasiveness. Here, we utilize a novel platform that combines mass spectrometry with an artificial intelligence-based data processing engine to interrogate the blood glycoproteome in melanoma patients before receiving ICI therapy. We identify 143 biomarkers that demonstrate a difference in expression between the patients who died within six months of starting ICI treatment and those who remained progression-free for three years. We then develop a glycoproteomic classifier that predicts benefit of immunotherapy (HR=2.7; p=0.026) and achieves a significant separation of patients in an independent cohort (HR=5.6; p=0.027). To understand how circulating glycoproteins may affect efficacy of treatment, we analyze the differences in glycosylation structure and discover a fucosylation signature in patients with shorter overall survival (OS). We then develop a fucosylation-based model that effectively stratifies patients (HR=3.5; p=0.0066). Together, our data demonstrate the utility of plasma glycoproteomics for biomarker discovery and prediction of ICI benefit in patients with metastatic melanoma and suggest that protein fucosylation may be a determinant of anti-tumor immunity.

9545 Background: Protein glycosylation is the most abundant and complex form of post-translational protein modification. Glycosylation profoundly affects protein structure, conformation, and function. The elucidation of the potential role of differential protein glycosylation as biomarkers has so far been limited by the technical complexity of generating and interpreting this information. We have recently established a novel, powerful platform that combines liquid chromatography/mass spectrometry with a proprietary artificial-intelligence-based data processing engine that allows, for the first time highly scalable interrogation of the glycoproteome. Methods: Using this platform, we interrogated 526 glycopeptide (GP) signatures derived from 75 serum proteins in pretreatment blood samples from a cohort of 205 individuals (66 females, 139 males, age range 24 to 97 years) with metastatic malignant melanoma treated either with nivolumab plus ipilimumab (95 patients) or pembrolizumab (110 patients) immune-checkpoint inhibitor (ICI) therapy. Results: In an optimized assay containing 27 glycopeptides and 20 non-glycosylated peptides, we identified 14 GPs with abundance differences at FDR q≤0.05 with regard to PFS. Using 40% of the cohort as a training set and selecting 12 glycopeptide and non-glycosylated peptide biomarker features of the 47 total by LASSO shrinkage, we created a multivariable-model-based classifier for PFS that yielded a hazard ratio (HR) for prediction of likely ICI benefit of 7.5 at p < 0.0001. This classifier was validated in the test set comprised of the held-out 60% of patients, yielding a HR of 4.7 at a similar p-value for separating patients likely benefiting from ICI therapy and those likely not benefiting from ICI therapy (50% PFS of 18 months vs. 3 months based on classifier score above/below cutoff). This classifier has a sensitivity of > 99% to predict likely ICI benefit, while still performing at a specificity of 26%, thus helping to safely reduce ultimately unnecessary and non-beneficial exposure to these agents of one in four who otherwise would unnecessarily be exposed to them. Conclusions: Our results indicate that glycoproteomics holds a strong promise as a predictor for checkpoint inhibitor treatment benefit that appears to significantly outperform other currently pursued biomarker approaches in this context.

e12545 Background: Breast cancer is the most common cancer among women worldwide.Traditional methods of cancer detection such as tissue biopsy are invasive, costly, time consuming and not amenable for repetition. As a result, minimally invasive liquid biopsies, especially blood-based biomarkers show potential value for breast cancer risk prediction and early detection. In this study, we investigated the use of serum glycoproteins circulating in blood to identify a panel of potential prognostic markers that may aid in predicting breast cancer in women. Methods: We applied a novel platform for characterizing blood glycoproteomic biomarkers, combining liquid-chromatography/mass spectrometry (LC-MS) with artificial intelligence/neural networks (AI-NN) to analyze serum samples from 279 breast cancer patients (median age 56 years, with stage 0-4 N’s: 1 / 83 / 114 / 56 / 25) and 102 healthy control samples (median age 52 years). A panel of 596 serum glycosylated and non-glycosylated peptides, representing 71 serum proteins, were analyzed. Age-adjusted differential expression analysis for 596 normalized biomarkers were performed to evaluate statistically significant differential abundances using an FDR q-value of 0.05 as a cutoff. Using the top differentially expressed markers as input, a LASSO penalized logistic regression model with 5-fold repeated cross validation was applied to identify the top biomarkers contributing to the separation between healthy controls and breast cancer patients. Results: We identified 243 out of 596 markers that were differentially expressed (FDR <<0.05) between breast cancer samples and healthy controls. Out of those, 11 markers were obtained as the top predictors in classifying breast cancer patients and healthy controls. The classification algorithm yielded an accuracy of 94% (95.9% sensitivity, 88.7% specificity) and an AUC of 0.983 on the training set. This classifier was validated on an independent test set with 30% of the subjects, yielding an accuracy of 93% (96.4% sensitivity, 83.9% specificity) and an AUC of 0.974. Test sensitivity was high across stages, at 96% / 90% / 95% / 90% in stages 1-4, respectively. Conclusions: Based on the results, we conclude that circulating glycoproteins in serum may be useful in screening applications in breast cancer, and strongly demonstrates the utility of glycoprotein profiles as a powerful non-invasive diagnostic tool.

No abstract