Multidimensional biomarker predicts disease control in response to immunotherapy in recurrent or metastatic head and neck squamous-cell carcinoma.

PURPOSE: Anti-PD-1 therapy provides clinical benefit in 40-50% of patients with relapsed and/or metastatic head and neck squamous cell carcinoma (RM-HNSCC). Selection of anti- PD-1 therapy is typically based on patient PD-L1 immunohistochemistry (IHC) which has low specificity for predicting disease control. Therefore, there is a critical need for a clinical biomarker that will predict clinical benefit to anti-PD-1 treatment with high specificity. METHODS: Clinical treatment and outcomes data for 103 RM-HNSCC patients were paired with RNA-sequencing data from formalin-fixed patient samples. Using logistic regression methods, we developed a novel biomarker classifier based on expression patterns in the tumor immune microenvironment to predict disease control with monotherapy PD-1 inhibitors (pembrolizumab and nivolumab). The performance of the biomarker was internally validated using out-of-bag methods. RESULTS: The biomarker significantly predicted disease control (65% in predicted non-progressors vs. 17% in predicted progressors, p < 0.001) and was significantly correlated with overall survival (OS; p = 0.004). In addition, the biomarker outperformed PD-L1 IHC across numerous metrics including sensitivity (0.79 vs 0.64, respectively; p = 0.005) and specificity (0.70 vs 0.61, respectively; p = 0.009). CONCLUSION: This novel assay uses tumor immune microenvironment expression data to predict disease control and OS with high sensitivity and specificity in patients with RM-HNSCC treated with anti-PD-1 monotherapy.

T cell subtype profiling measures exhaustion and predicts anti-PD-1 response.

Anti-PD-1 therapy can provide long, durable benefit to a fraction of patients. The on-label PD-L1 test, however, does not accurately predict response. To build a better biomarker, we created a method called T Cell Subtype Profiling (TCSP) that characterizes the abundance of T cell subtypes (TCSs) in FFPE specimens using five RNA models. These TCS RNA models are created using functional methods, and robustly discriminate between naïve, activated, exhausted, effector memory, and central memory TCSs, without the reliance on non-specific, classical markers. TCSP is analytically valid and corroborates associations between TCSs and clinical outcomes. Multianalyte biomarkers based on TCS estimates predicted response to anti-PD-1 therapy in three different cancers and outperformed the indicated PD-L1 test, as well as Tumor Mutational Burden. Given the utility of TCSP, we investigated the abundance of TCSs in TCGA cancers and created a portal to enable researchers to discover other TCSP-based biomarkers.

Analytical Performance of an Immunoprofiling Assay Based on RNA Models.

As immuno-oncology drugs grow more popular in the treatment of cancer, better methods are needed to quantify the tumor immune cell component to determine which patients are most likely to benefit from treatment. Methods such as flow cytometry can accurately assess the composition of infiltrating immune cells; however, they show limited use in formalin-fixed, paraffin-embedded (FFPE) specimens. This article describes a novel hybrid-capture RNA sequencing assay, ImmunoPrism, that estimates the relative percentage abundance of eight immune cell types in FFPE solid tumors. Immune health expression models were generated using machine learning methods and used to uniquely identify each immune cell type using the most discriminatively expressed genes. The analytical performance of the assay was assessed using 101 libraries from 40 FFPE and 32 fresh-frozen samples. With defined samples, ImmunoPrism had a precision of ±2.72%, a total error of 2.75%, and a strong correlation (r2 = 0.81; P < 0.001) to flow cytometry. ImmunoPrism had similar performance in dissociated tumor cell samples (total error of 8.12%) and correlated strongly with immunohistochemistry (CD8: r2 = 0.83; P < 0.001) in FFPE samples. Other performance metrics were determined, including limit of detection, reportable range, and reproducibility. The approach used for analytical validation is shared here so that it may serve as a helpful framework for other laboratories when validating future complex RNA-based assays.

Automated condition-invariable neurite segmentation and synapse classification using textural analysis-based machine-learning algorithms.

High-resolution live-cell imaging studies of neuronal structure and function are characterized by large variability in image acquisition conditions due to background and sample variations as well as low signal-to-noise ratio. The lack of automated image analysis tools that can be generalized for varying image acquisition conditions represents one of the main challenges in the field of biomedical image analysis. Specifically, segmentation of the axonal/dendritic arborizations in brightfield or fluorescence imaging studies is extremely labor-intensive and still performed mostly manually. Here we describe a fully automated machine-learning approach based on textural analysis algorithms for segmenting neuronal arborizations in high-resolution brightfield images of live cultured neurons. We compare performance of our algorithm to manual segmentation and show that it combines 90% accuracy, with similarly high levels of specificity and sensitivity. Moreover, the algorithm maintains high performance levels under a wide range of image acquisition conditions indicating that it is largely condition-invariable. We further describe an application of this algorithm to fully automated synapse localization and classification in fluorescence imaging studies based on synaptic activity. Textural analysis-based machine-learning approach thus offers a high performance condition-invariable tool for automated neurite segmentation.