In lung adenocarcinoma (LUAD), tumors are graded along a spectrum of cell differentiation states, with "low grade" or "well differentiated" tumors being associated with better overall survival. Understanding what drives tumors from a low grade to a higher grade is an area of active investigation in an effort to identify new therapeutic targets. Typically, these investigations use 'omics techinques on bulk tumor samples from biopsies. However, tumors are heterogenous in nature and these bulk analyses provide a snapshot of the average cell state within the tumor. Identifying drivers of tumor progression would be greatly empowered by comparing regions of different grades within individual tumors.

Using our machine learning model for grading LUAD (GLASS-AI) we can reveal the heterogeneity of tumor grade throughout a tissue section. We can use the grades provided by GLASS-AI to select tumor regions for analysis using our spatial multi-omics workflow. 

In this example image, comparing regions 1 and 2 would allow us to uncover the changes in gene activity, protein expression, and metabolic activity that are associated with the progression from grade 2 (blue) to grade 3 (yellow) in a single tumor. By performing similar comparisons across multiple tumors and specimens, we can identify molecular signatures unique to each grade of LUAD. 

We can also compare the multi-omics results among tumor regions of a single grade from multiple tumors, such as regions 2 and 3. These comparisons will not only aid in the definition of concise molecular signatures for each grade but can also be used to develop a progression risk score that could supplement the existing tumor grading systems, including GLASS-AI.