For over 100 years, the traditional tools of pathology, such as tissue-marking dyes (e.g. the H&E stain) have been used to study the disorganization and dysfunction of cells within tissues. This has represented a principal diagnostic and prognostic tool in cancer. However, in the last 5 years, new technologies have promised to revolutionize histopathology, with Spatial Transcriptomics technologies allowing us to measure gene expression directly in pathology-stained tissue sections. In parallel with these developments, Artificial Intelligence (AI) applied to histopathology tissue images now approaches pathologist level performance in cell type identification. However, these new technologies still have severe limitations, with Spatial Transcriptomics suffering difficulties distinguishing transcriptionally similar cell types, and AI-based pathology tools often performing poorly on real world out-of-batch test datasets. Thus, century-old techniques still represent standard-of-care in most areas of clinical cancer diagnostics and prognostics. Here, we present a new frontier in digital pathology: describing a conceptually novel computational methodology, based on Bayesian probabilistic modelling, that allows Spatial Transcriptomics data to be leveraged together with the output of deep learning-based AI used to computationally annotate H&E-stained sections of the same tumor. By leveraging cell-type annotations from multiple independent pathologists, we show that this integrated methodology achieves better performance than any given pathologist’s manual tissue annotation in the task of identifying regions of immune cell infiltration in breast cancer, and easily outperforms either technology alone. We also show that on a subset of histopathology slides examined, the methodology can identify regions of clinically relevant immune cell infiltration that were missed entirely by an initial pathologist’s manual annotation. While this use case has clear diagnostic and prognostic value in cancer (e.g. predicting response to immunotherapy), our methodology is generalizable to any type of pathology images and also has broad applications in spatial transcriptomics data analytics, where most applications (such as identifying cell-cell interactions) rely on correct cell type annotations having been established a priori. We anticipate that this work will spur many follow-up studies, including new computational innovations building on the approach. The work sets the stage for better-than-pathologist performance in other cell-type annotation tasks, with relevant applications in diagnostics and prognostics across almost all cancers. Citation Format: Asif Zubair, Rich Chapple, Sivaraman Natarajan, William C. Wright, Min Pan, Hyeong-Min Lee, Heather Tillman, John Easton, Paul Geeleher. Jointly leveraging spatial transcriptomics and deep learning models for image annotation achieves better-than-pathologist performance in cell type identification in tumors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 456.

Neuroblastoma is a highly heterogeneous disease not only in the clinical presentation of individual patients, but also in the cellular composition of any given tumor. Insights into this diversity have only recently been enabled due to advancements in single cell technologies, which have facilitated investigation of this disease at unprecedented resolution and detail. Coinciding with the growing number of scRNA-seq technologies, so too are the number of single cell datasets encompassing neuroblastoma patients across several institutions. However, due to the rarity of the affliction and sample access, the cohort pool in each aforementioned scRNA-seq study is limited to a reduced representation of the spectrum of disease classifications, which limits the ability of any single study to draw conclusions about neuroblastoma as a whole. Moreover, inconsistencies in data acquisition and analytical approaches across these studies have led to diverging interpretations. As such, we decided to amass the entirety of publicly available neuroblastoma scRNA-seq studies, representing a more comprehensive cross-section of patient presentations, towards the goal of conducting an exhaustive meta-analysis of the underlying data. To this end, we have implemented a generalizable non-negative matrix factorization (NMF)-based framework targeted at discovering conserved gene expression programs in malignant neuroblastoma as well as the supporting microenvironment. Using graph-based network analyses for classification of gene expression programs, we have identified conserved signatures of malignant and non-tumor cell types in neuroblastoma. In addition to defining the landscape of expression programs in human neuroblastoma patients, we have also utilized the NMF analysis to assess the alignment of several preclinical models to human signatures. We have identified gene expression programs that align to malignant human expression programs as well as signatures more closely related to non-tumor cell types. These include previously characterized divergent mesenchymal and adrenergic programs, as well as undescribed liver/metabolic, neuronal, and glial signatures. When considering the affinity of neuroblastoma models to malignant human profiles we observed specific agreement between certain preclinical signatures and subtype classifications found in patient samples. Careful consideration of these results will allow researchers to guide preclinical studies by cross-referencing neuroblastoma models of interest with patient profiles. Overall, we characterize the most updated view of the landscape of neuroblastoma by documenting the full repertoire of gene expression programs across patient and preclinical models. Citation Format: Richard Chapple, Charlie Wright, Min Pan, Paul Geeleher. Meta-analysis of neuroblastoma single cell RNA-seq datasets identifies conserved and divergent gene expression programs across human and preclinical models [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4075.

Gene set analysis (GSA) remains a common step in genome-scale studies because it can reveal insights that are not apparent from results obtained for individual genes. Many different computational tools are applied for GSA, which may be sensitive to different types of signals; however, most methods test whether there are differences in the distribution of the effect of some experimental condition between genes in gene sets of interest. We have developed a unifying framework for GSA that first fits effect size distributions, and then tests for differences in these distributions between gene sets. These differences can be in the proportions of genes that are perturbed or in the sign or size of the effects. Inspired by statistical meta-analysis, we take into account the uncertainty in effect size estimates to reduce the influence of genes with greater uncertainty in effect size estimate on distribution parameters. We demonstrate, using simulation and by application to real data, that this approach provides significant gains in performance over existing methods. Furthermore, the statistical tests carried out are defined in terms of effect sizes, rather than the results of prior statistical tests measuring these changes, which leads to improved interpretability and greater robustness to variation in sample sizes. We also show that the approach naturally suggests alternative test types that are not usually considered in GSA; it can, for example, be applied to identify differences in effect size distributions between sample subgroups in a gene set of interest. Applying this approach to an analysis of gene expression changes between matched colon tumour and normal samples, we found several gene sets that showed distinct behaviour in patient subgroups with different prognoses. These may help to explain the clinical differences that have been reported between these patient groups.

Drug combinations are the basis of treatment for modern diseases but arriving at successful combination therapies is fraught with challenges. Decades ago, the limited number of drugs represented a tractable candidate list from which to design combination experiments. However, the current pool of single-agent drugs to potentially combine is far too large to brute-force screen, and purely computational predictions have performed poorly. Suitable screening methods are needed, but the design of experimental approaches has proven to be highly complex; researchers need to carefully balance many variables such as appropriate drug concentration ranges, number of doses, inclusion of replicates, and throughput. Perhaps the most significant obstacle facing these studies is the approach to data analysis, where conflicting definitions of synergy and unintuitive metrics serve to confuse researchers and render largely uninterpretable results. Collectively, these challenges hamper the progress of drug combination research and ultimately translational impact. To overcome these limitations, we have developed a fully self-contained framework to handle both the experimental design and analysis of drug combination experiments. Our method, called Combocat, provides a straightforward way to test and analyze any number of drug combinations and samples, and is suitable for high-throughput. Combocat provides a high-resolution of concentration combinations compared to most current approaches. This is automated by common instruments and uses scripts included within our protocol. Through careful template design, we were able to include 3 replicates of each 10x10 matrix, single-agent drugs, and controls - all within a single 384-well plate. We found our method to work robustly with varying sample types (Human cancer, bacteria, fungi), and readouts. After data generation, files can simply be dragged into our Combocat analysis tool directly. We provide a free, web-based software suite to fully automate the analysis after data collection. The Combocat web tool is intuitive and facilitates interactive exploration of synergy. It also provides a rich array of information such as dose-response curves, IC50 values, synergy matrices, ranked hit plots, and more. Data normalization, synergy algorithms, scoring functions, and other complex calculations are run swiftly and automatically in the background with no need for user input. Notably, we employ statistical testing by taking advantage of experimental replicates, which is a feature we found lacking in most methods. We use a well-documented synergy metric but also decided to formulate our own Combocat score which considers statistical measurements and assay quality. The Combocat score provides an easy interpretation of results and facilitates quick identification of top hits. Collectively, our platform will be used to enhance and expedite the selection of effective drug combinations. Citation Format: William C. Wright, Min Pan, Hyeong-Min Lee, Gregory A. Phelps, Jonathan Low, Duane Currier, Richard E. Lee, Taosheng Chen, Paul Geeleher. Combocat: A high-throughput framework for drug combination studies [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1907.