A computational framework to address tumor microenvironment of pancreatic ductal adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is a deadly malignancy, with a 5-year survival rate of only 5%. The notorious intra- and inter-tumoral heterogeneity of PDAC have been a significant obstacle in elucidating molecular and cellular mechanisms for developing treatment, disease progression, metastasis, and recurrence. Therefore, deciphering the heterogeneity of tumor cells and the tumor microenvironment (TME) is essential to improve therapeutic strategies for PDAC.
Tumor heterogeneity has been studied commonly by characterizing tumor subtypes with distinct gene expression profiles. However, classical gene expression profiling provides only a little information on the TME, despite its
critical contribution to disease progression. In particular, the immune components in the TME and their interaction with cancer cells might enable immune evasion, rendering failure of immunotherapy. Singlecell technologies, such as single-cell RNA-seq and Mass cytometry, have recently emerged as effective tools for detailed TME profiling. Unfortunately, these single-cell technologies come with 1) high-cost, 2) low signal-to-noise ratio, and 3) time-consuming, laborious manual processing of fresh tissue.

We propose to interrogate the intra-tumoral heterogeneity of PDAC by characterizing 1) TME cell type fractions, 2) detailed functional characteristics of each cell type, and 3) cellular interactions. We will build upon BLADE (Bayesian LognormAl DEconvolution), which infers proportions of cell types and cell-typespecific gene expression profiles from bulk RNA-seq data. By providing a cell-type-specific gene expression profile in PDAC, BLADE can be tailored to study PDAC TME. Furthermore, a detailed functional characterization of TME cell types is possible by extending BLADE to estimate cell-typespecific multi-omics profiles. Moreover, we can characterize interactions between tumor and TME cells
by capturing the spatial information of cells within the tumor. This comprehensive TME profiling of multiple large-scale PDAC cohorts may allow us to delineate key TME cell types and their interactions in treatment response.

We aim to gain insights into the significance of the PDAC tumor microenvironment in the prognosis and treatment response by characterizing TME profiles in a large number of PDAC samples at an unprecedented level of comprehensiveness.
Functional characterization of individual TME cell types associated with a distinct patient outcome may lead to novel non-tumor therapeutic targets and targetable key signaling pathways. Furthermore, characterization of interactions between TME cells may elucidate the dependency of tumor cells to surrounding normal cells for the disease progression and resistance to treatments. In particular, this may shed light on immune-evasion mechanisms, such as the key pathways modulating PD-1 and PD-L1.
Due to the complementary nature of techniques used in the study, establishing a computational framework must be the central task to gain robust knowledge by
integrating information from each data type. Collectively, we hypothesize that this comprehensive multilayered characterization of the PDAC microenvironment will lead to the identification of potential vulnerability of PDAC that can be leveraged to fight against this devastating disease.

WP1. We will tailor BLADE to study the PDAC TME. Based on the detailed characterization of systemic immune profiles, we will elucidate the associations between the tumor microenvironment and systemic immune profiles.
WP2. We will add an extra layer to BLADE, called multi-layered BLADE, to functionally characterize each PDAC TME cell type.
WP3.We will further characterize PDAC TME cell types by their cellular interplays within the tumor tissues.
Expected outcome:
A tailored deconvolution tool to study PDAC TME will be delivered, which can also be used in future studies. The notorious stromal component and heterogeneity of PDAC make it ideal for establishing such tools to be applied in other cancer types. The functionally characterized PDAC TME cell types and interactions profiled in many PDAC tumors may guide us towards specific vulnerabilities of PDAC tumors that can improve disease management. In particular, we expect that TME cell types characterized in this study can be used in future PDAC studies involving transcriptome profiling. The detailed characterization of PDAC TME cell types in large-scale PDAC samples may shed light on key cell types underlining 1) the different biology between molecular subtypes previously described and 2) distinct survival outcomes of the patients. Besides, our spatial interaction layer elucidates the importance of spatial characteristics in developing TME characteristics-based biomarkers. Finally, systemic immune characterization in WP1 could lead to a liquid-biopsy-based diagnostic tool to monitor PDAC immune microenvironment.