Citations

Abstract Pediatric Acute Myeloid Leukemia (AML) with a FLT3 internal tandem duplication (FLT3-ITD) is a challenging disease due to poor outcomes in many patients. The 4-year progression-free survival is still only 31%. Current biomarkers are insufficient to predict why certain patients with FLT3-ITD AML relapse and others do not. The development of prognostic biomarkers in FLT3-ITD pediatric AML may help improve the outcomes and management of these patients. We acquired a panel of 37 diagnostic samples with 18 samples from FLT3-ITD pediatric AML patients with poor outcomes (relapse within 3 years) and 19 from favorable outcome patients that did not exhibit relapse for at least 5 years. Next, we performed FACS to isolate CD34+CD38dim and CD34+CD38+ cells. These two populations were then sequenced using 10x Genomics 3’ end single-cell RNA sequencing (scRNAseq). Using this scRNAseq dataset comprising over 250k single cells, we first investigated if the frequency of specific RNA clusters of AML cells may predict patient outcome. We found 5 clusters had significantly different frequencies between patients with favorable and poor outcomes (p < 0.05). Using these clusters, we performed CIBERSORTx deconvolution on publicly available FLT3-ITD AML bulk RNAseq samples from TCGA (n = 36), BeatAML (n = 47), and the TARGET project (ntrain = 72; ntest = 26) to identify the inferred cluster composition present in each AML patient from these datasets. Next we built LASSO prediction models using the inferred scRNAseq clusters by CIBERSORTx and incorporating clinical features including leukemic blast percentage, bone marrow leukemic blast percentage, and WBC at diagnosis. We found these prediction models to be highly specific as a prognostic biomarker for pediatric AML (p << 0.01). We were unable to train a model directly using genes from the publicly available bulk RNAseq data for pediatric FLT3-ITD AML, highlighting the relevance of using our deconvolution based biomarker. Between the high (> median) and low (< median) LASSO score groups, there was ~50% difference in the number of patients relapsing after 4 years. When we included this LASSO score in a multivariate cox regression (likelihood ratio p value = 0.007), the LASSO score (HR = 6.8, p = 0.026, CI = 1.26 - 37.28) showed additional predictive ability when including FLT3-ITD allelic ratio (HR = 1.83, p = 0.38, CI = 0.47 - 7.15) and MRD (HR = 3.95, p = 0.021 CI = 0.051 - 0.79). These results demonstrate that incorporating rare AML cell populations into a prognostic biomarker could add substantial clinical information to existing biomarkers. This indicates single-cell derived prognostic models can and should be explored further for pediatric FLT3-ITD AML and demonstrates a platform that could be used for other malignancies, including non-hematologic tumors, to create cell-type biomarkers. Citation Format: Robert Schauner, Zachary Jackson, Nethrie Idippily, Grace Lee, Sheela Karunanithi, Shivaprasaad Manjappa, Tae Hyun Hwang, David Wald. New platform for novel biomarker discovery using single-cell RNA sequencing on pediatric acute myeloid leukemia samples [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 4080.

Abstract Histological distribution and density of tumor infiltrating lymphocytes (TILs) are known to be informative to cancer patients’ outcome. However, due to the heterogeneity and huge size of digitized pathology slides, it has been a challenging problem to quantify TILs-related features in whole slide images (WSI). This study aimed to explore the impact of spatial distribution of TILs quantified by deep learning (DL) approaches based on digitalized WSI stained with hematoxylin and eosin in patients with colorectal cancer (CRC). The cancer and TILs regions are first separately identified by deep learning models in the WSI. The TILs densities at tumor invasive margins are then quantified by image analysis algorithms and used for prognosis of progression-free survival (PFS) in CRC patients. The prognostic impact of spatial distributions of TILs in CRC patients was explored in the Yonsei cohort (n=180) and validated in the TCGA cohort (n=268). Two experienced pathologists manually measured TILs at the most invasive margin as 0-3 by the Klintrup-Mäkinen (KM) grading method and compared to DL approaches. On multivariate analysis, TILs densities within 200µm of the invasive margin (f_im200) was remained as the most significant prognostic factor for PFS (HR 0.004 [95% CI, 0.0001-0.1502], p=.002) in the Yonsei cohort. On multivariate analysis using the TCGA dataset, f_im200 retained prognostic significance for PFS (HR 0.031, [95% CI 0.001-0.645], p=.024). Inter-observer agreement of manual KM grading was insignificant in both cohorts. The survival analysis by pathologists’ manual KM grading showed statistically significant different PFS from the TCGA cohort, but not the Yonsei cohort. In summary, automated quantification of TILs at the invasive margin showed a prognostic utility to predict PFS, and could provide reproducible TILs density measurement in CRC patients. Citation Format: Hongming Xu, Yoon Jin Cha, Jean Clemenceau, Jinhwan Choi, Sung Hak Lee, Jeonghyun Kang, Tae Hyun Hwang. Spatial analysis of tumor infiltrating lymphocytes in histological images using deep learning predicts progression-free survival in colorectal cancer [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 642.

Abstract Background: Microsatellite Instability (MSI) is a prognostic and predictive biomarker which can guide treatments including immunotherapy for patients with gastrointestinal cancer. Universal MSI testing would benefit to patients to receive a better therapy, but many patients remain untested. To deliver broadly accessible MSI testing, deep learning models have demonstrated their feasibility for MSI prediction using H&E-stained whole-slide images (WSIs). However, to be deployed into clinical routine care, prediction models should be validated in datasets from multicenter and multiethnic groups. In addition, these models should provide uncertainty of prediction to help clinicians to make informed decisions. Method: We develop a MSI prediction model using WSIs based on Deep Gaussian process (DGP), a Bayesian model being able to model the uncertainty in prediction. We implement a DGP model in the transfer learning, where a WPI is decomposed into multiple non-overlap image patches which are converted to feature vectors by a pretrained convolutional neural network (CNN). Then, the DGP model makes a prediction for an image by averaging score function values at all the feature vectors in the image in weakly supervised learning (WSL). To test our method, we collected H&E stained colorectal/stomach cancer slides (n=1,619) from multi-institutions and multiethnic groups, including TCGA Colorectal (CRC; n=351) and Stomach Adenocarcinoma (STAD; n=174) and six datasets from tertiary hospitals in Korea (i.e., Yonsei-1 (n=174), Yonsei-1-remade (n=146), Yonsei-2 (n=95), St. Mary-1 (n=48), St. Mary-2 (n=50), and Yonsei-CLASSIC trial (n=581)). We also compare our method to multiple deep learning methods, including Densenet, Resnet, Shufflenet, Googlenet, Squeezenet and two WSL models, MIL (Campanell et al., 2019) and CLAM (Lu et al., 2021). Result: In each cancer type, we train each model using TCGA dataset as discovery cohort and tested on the rest of datasets as validation cohort. We use the area under of roc curve (AUC) for the evaluation metric. For the CRC cohorts, our DGP method achieves overall best performance with 0.81 (Yonsei-1), 0.82 (Yonsei-1-remade), 0.89 (Yonsei-2), 0.85 (St. Mary-1) 0.75 (St. Mary-2) AUCs compared to the state-of-the art methods. For the STAD cohorts, our method also achieves overall best performance with 0.74 AUC (Yonsei-classic). In addition, incorporating the uncertainty in prediction measured by our method improves the model’s performance to predict MSI. Specifically, removing cases with prediction results with high uncertainty, which could lead false positive prediction, significantly improves MSI prediction performance. Taken together, these results demonstrate the robustness and generalisability of the DGP model for MSI prediction across multicenter and multiethnic datasets. Citation Format: Sunho Park, Hongming Xu, Sung Hak Lee, Jeonghyun Kang, Tae Hyun Hwang. Deep Gaussian process with uncertainty estimation Improves microsatellite instability prediction based on whole slide image: A retrospective multicenter and multiethnic cohort study [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 5012.

Abstract Background: Gastric carcinogenesis has been found to be associated with intragastric microbiome, gastric mucosal gene expression, and immune cells comprising the tumor microenvironment separately in independent studies. However, to date, there has been no multi-omic investigations analyzing these biological traits simultaneously. Thus, there is a paucity in our understanding of the intricate relationship between intragastric microbes and the host in gastric disease. To address this gap, we aimed to comprehensively investigate the microbiome, host gene expression, and immune cells in the stomach across a range of different states along the gastric carcinogenesis pathway. Methods: Gastric biopsy samples were obtained from healthy individuals or patients with gastritis or gastric cancer (biopsy cohort, n=30). Surgical samples (of cancer and adjacent severe gastritis tissues) were obtained from gastric cancer patients (surgery cohort, n=40). Using 16S rRNA gene sequencing, RNA-seq, and cell-type enrichment analysis of each participant’s biospecimens, we identified associations between the microbiome and host transcriptome or inferred immune cell types. Results: Microbiome composition and gene expression patterns showed significant differences between disease states (healthy, gastritis, or cancer). Expression of ‘Cell cycle’ pathway genes was enriched in gastric cancer in both biopsy and surgery cohorts. In the biopsy cohort, Helicobacteraceae, which was abundant in gastritis, was highly correlated with FAM3D, which was previously found to be involved in gastrointestinal inflammation. In the surgery cohort, Lachnospiraceae, which was abundant in gastric cancer, was highly correlated with UBD, which is known to regulate mitosis and reduce cell cycle time. Cell-type enrichment analysis revealed lower B cell infiltration in gastric cancer than in gastritis (both cohorts). Tissue-infiltrating B cells were associated with Helicobacteraceae. Conclusions: Associations between the intragastric microbiome, human gene expression, and tissue-infiltrating immune cells are unique to gastric disease state. The findings herein may aid the discovery of novel biomarkers or therapeutic targets of gastric cancer. Citation Format: Chan Hyuk Park, Changjin Hong, A-reum Lee, Jaeyun Sung, Tae Hyun Hwang. Integrated analysis of intragastric microbiome and human gene expression uncovers genetic, microbial, and immunological associations in gastritis and gastric cancer [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 5013.

Abstract Background: The tumor microenvironment is a biological environment comprising of heterogeneous cellular components, such as tumor, immune, and stromal cells. Stromal cells in the tumor microenvironment have been shown that they play an important role in cancer development, but how they are linked to patient prognosis, chemotherapy response, and immunotherapy resistance is not fully investigated yet. The present study sought to investigate the stroma composition in the tumor microenvironment by using spatial analysis and deep learning. Method: We analyzed stroma composition in Whole Slide Images (WSIs) of hematoxylin and eosin-stained formalin-fixed paraffin-embedded (FFPE) sections from 4 patient cohorts: St. Mary (37 patients), Yonsei-1 (622 patients), Yonsei-2 (66 patients), and TCGA-STAD (359 patients). St. Mary was collected from St. Mary Hospital, Seoul, Republic of Korea. Yonsei-1 and Yonsei-2 were collected from Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea. Each WSI was tessellated into 256x256 um at 40X magnification and normalized using Macenko and Reinhard normalization methods to reduce the color variability of the stained tissue images. Two deep learning models (ResNet and U-Net) were used to create prediction models using Kather’s tumor dataset and a crowdsourcing stroma dataset. Models were trained over 100 epochs, using the Adam optimizer, 1e-4 learning rate, a binary loss function, and L2 regularization. Trained models were used to predict WSIs including tumor, stroma, tumor-infiltrating lymphocytes (TILs), Microsatellite instability. Spatial analysis was conducted by using Getis-Ord-Gi*, a local statistic that identifies a statistically significant hot-spot region by providing relatedness with its neighbors. Tumor hot-spot regions were computed based on z-scores in the predicted WSIs and used for understanding the spatial variation of cancer cells, recognizing statistically significant regions and their spatial relatedness. Results: In Yonsei-1, stroma composition was significantly higher in the chemotherapy high-risk group (p-value 2.9e-09). The proportion of stroma composition is significantly different with the risk groups (p-value 0.074). In TCGA-STAD, the proportion of stroma composition is significantly different with ACTA2 and Microsatellite instability (p-value 0.028). In St. Mary, Yonsei-1, and Yonsei-2, stroma composition was significantly higher in the immunotherapy treatment non-response group (p-value 0.001). Conclusion:These findings suggest that stroma composition in the tumor microenvironment is strongly associated with patient prognosis, chemotherapy response, and immunotherapy resistance on stomach cancer. Citation Format: Sanghoon Lee, Isaiah Pressman, Hongming Xu, Tae Hyun Hwang. Spatial image analysis on stroma composition linking to patient prognosis, chemotherapy response, and immunotherapy resistance [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 5044.

Abstract Chimeric antigen receptor T-cell (CAR-T cell) therapy directed at CD19 produces durable remissions in the treatment of relapsed/refractory non-Hodgkin lymphoma (NHL). Nonetheless, many patients receiving CD19 CAR-T cells fail to respond for unknown reasons. To reveal changes in 4-1BB–based CD19 CAR-T cells and identify biomarkers of response, we used single-cell RNA sequencing and protein surface marker profiling of patient CAR-T cells pre- and postinfusion into patients with NHL. At the transcriptional and protein levels, we note the evolution of CAR-T cells toward a nonproliferative, highly differentiated, and exhausted state, with an enriched exhaustion profile in CAR-T cells of patients with poor response marked by TIGIT expression. Utilizing in vitro and in vivo studies, we demonstrate that TIGIT blockade alone improves the antitumor function of CAR-T cells. Altogether, we provide evidence of CAR-T cell dysfunction marked by TIGIT expression driving a poor response in patients with NHL. Significance: This is the first study investigating the mechanisms linked to CAR-T patient responses based on the sequential analysis of manufactured and infused CAR-T cells using single-cell RNA and protein expression data. Furthermore, our findings are the first to demonstrate an improvement of CAR-T cell efficacy with TIGIT inhibition alone. This article is highlighted in the In This Issue feature, p. 1825

Abstract This study aimed to explore the prognostic impact of spatial distribution of tumor‐infiltrating lymphocytes (TILs) quantified by deep learning (DL) approaches based on digitalized whole‐slide images stained with hematoxylin and eosin in patients with colorectal cancer (CRC). The prognostic impact of spatial distributions of TILs in patients with CRC was explored in the Yonsei cohort ( n = 180) and validated in The Cancer Genome Atlas (TCGA) cohort ( n = 268). Two experienced pathologists manually measured TILs at the most invasive margin (IM) as 0–3 by the Klintrup–Mäkinen (KM) grading method and this was compared to DL approaches. Inter‐rater agreement for TILs was measured using Cohen's kappa coefficient. On multivariate analysis of spatial TIL features derived by DL approaches and clinicopathological variables including tumor stage, microsatellite instability, and KRAS mutation, TIL densities within 200 μm of the IM (f_im200) remained the most significant prognostic factor for progression‐free survival (PFS) (hazard ratio [HR] 0.004 [95% confidence interval, CI, 0.0001–0.15], p = 0.0028) in the Yonsei cohort. On multivariate analysis using the TCGA dataset, f_im200 retained prognostic significance for PFS (HR 0.031 [95% CI 0.001–0.645], p = 0.024). Inter‐rater agreement of manual KM grading was insignificant in the Yonsei ( κ = 0.109) and the TCGA ( κ = 0.121) cohorts. The survival analysis based on KM grading showed statistically significant different PFS in the TCGA cohort, but not the Yonsei cohort. Automatic quantification of TILs at the IM based on DL approaches shows prognostic utility to predict PFS, and could provide robust and reproducible TIL density measurement in patients with CRC.

AbstractMotivation: A central problem in biomarker discovery from large-scale gene expression or single nucleotide polymorphism (SNP) data is the computational challenge of taking into account the dependence among all the features. Methods that ignore the dependence usually identify non-reproducible biomarkers across independent datasets. We introduce a new graph-based semi-supervised feature classification algorithm to identify discriminative disease markers by learning on bipartite graphs. Our algorithm directly classifies the feature nodes in a bipartite graph as positive, negative or neutral with network propagation to capture the dependence among both samples and features (clinical and genetic variables) by exploring bi-cluster structures in a graph. Two features of our algorithm are: (1) our algorithm can find a global optimal labeling to capture the dependence among all the features and thus, generates highly reproducible results across independent microarray or other high-thoughput datasets, (2) our algorithm is capable of handling hundreds of thousands of features and thus, is particularly useful for biomarker identification from high-throughput gene expression and SNP data. In addition, although designed for classifying features, our algorithm can also simultaneously classify test samples for disease prognosis/diagnosis.Results: We applied the network propagation algorithm to study three large-scale breast cancer datasets. Our algorithm achieved competitive classification performance compared with SVMs and other baseline methods, and identified several markers with clinical or biological relevance with the disease. More importantly, our algorithm also identified highly reproducible marker genes and enriched functions from the independent datasets.Availability: Supplementary results and source code are available at http://compbio.cs.umn.edu/Feature_Class.Contact: kuang@cs.umn.eduSupplementary information: Supplementary data are available at Bioinformatics online.