Chapter 7 240 in KRAS mutated NSCLC tumors39. Furthermore, high levels of Beta-2-microglobulin (B2M) mRNA, a component of MHC class I molecules, in baseline tumor samples have been correlated with an enhanced response to PD-1 blockade monotherapy in melanoma40. Lastly, the expansion of immune-inhibitory populations within the TME, such as myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), has been associated with poor patient survival41–43. Chapter 4; A dysfunctional T cell gene signature that is more suitable for routine clinical diagnostics In chapter 2, we reported a reliable automated method for digital quantification of PD-1T TILs using IHC, based on an approach from earlier work7. However, IHC staining results may vary, primarily due to preanalytical factors such as tissue and slide preparation, and analytical factors such as antibody validation and the use of different staining platforms. Furthermore, a substantial degree of user involvement is necessary for the digital quantification of PD-1T TILs, particularly as the delineation of tumor regions requires manual annotation. Computational pathology, using artificial intelligence (AI) techniques, holds the potential to enhance standardization across different medical centers. AI can perform complex visual pattern recognition by compiling subtle visual cues that are related to cell counts, cellular morphology, textures, and spatial patterns. For instance, the digital quantification of PD-L1 expression levels can be accomplished through IHC using machine learning techniques44,45. However, AI-based biomarkers come with several limitations. For example, it requires high-quality training data because the presence of artefactual images can introduce noise, necessitating a substantial number of images for the model to attain satisfactory performance. Furthermore, multiple datasets are needed to accurately represent diverse real-world patient populations, which can vary significantly between different hospitals. Therefore, data standardization and quality control systems are essential before application in routine clinical diagnostics46. An ideal platform for biomarker assessment in routine diagnostic applications should deliver sensitive, specific, and reproducible results while requiring minimal complexity for hands-on laboratory work and data analysis. The Nanostring nCounter platform has demonstrated its ability to meet this criteria, even when dealing with low-quality RNA samples obtained from FFPE samples47. This technique hybridizes fluorescent barcodes directly to specific nucleic acid sequences, enabling individual counting without the need of amplification, and it offers enhanced precision compared to microarray and RNA sequencing47,48. Importantly, the Nanostring nCounter platform has previously demonstrated its clinical applicability49.
RkJQdWJsaXNoZXIy MTk4NDMw