Pathology has undergone a remarkable transition over the past few years, shifting from traditional visual methods to computational pathology. Despite major advancements in the field, Mark Gustavson, Senior Director of Translational Medicine, Oncology R&D, AstraZeneca, claimed that the industry has yet to develop an improved companion diagnostic using computational pathology methods.
Gustavson and his team are developing a method that uses IHC to quantify protein expression and understand spatial heterogeneity. IHC has contributed to the approval of several revolutionary drugs including pembrolizumab and durvalumab. Traditional IHC scoring is done by a pathologist meaning that it is somewhat subjective and semi-quantitative since the pathologist passes judgment on what category to place patients in. Additionally, traditional IHC typically fails to quantify spatial heterogeneity or low-level protein expression. Meanwhile, computational biology is truly quantitative and objective.
AstraZeneca developed a method called QCS (quantitative continuous score) that uses artificial intelligence and deep learning to quantify protein expression in tissue samples, providing continuous data rather than categorical data. This method aims to improve the accuracy and sensitivity of biomarker analysis. Gustavson elaborated on this by saying that the QCS method is very applicable to ADCs because it supports researchers in understanding spatial heterogeneity and the distribution of biomarkers within tumours.
Gustavson focused on the relationship between TDX-d and HER2 expression. He also homed in on the influence of bystander activity on ADCs. The bystander effect refers to the ability of ADCs to kill not only targeted tumour cells but also neighbouring, antigen-negative cells, regardless of their antigen expression status.  A spatial proximity score measured how close HER2 negative cells are to the HER2 positive ones, which is important for ADCs with bystander effects.  Experiments showed that traditional HER2 scoring misses patients with low HER2 expression who may still benefit from TDX-d, so Gustavson wanted to tackle this using QCS.
QCS identified more patients as HER2-positive compared to traditional methods, with similar or better efficacy outcomes. Gstavson commented: “When we use all patients again all expression levels, 72 patients were identified as HER2 positive with QCS, we identified 120 patients that were QCS positive with the same efficacy.”
AstraZeneca is carrying out ongoing efforts to refine and validate the QCS method across different cancer types. The QCS method was validated across independent datasets. Researchers are looking into its potential integration with other technologies like mass spectrometry for biomarker analysis.
