Functional interrogation of twenty type 2 diabetes-associated genes using isogenic human embryonic stem cell-derived β-like cells

Genetic studies have identified numerous loci associated with type 2 diabetes (T2D), but the functional roles of many loci remain unexplored. Here, we engineered isogenic knockout human embryonic stem cell lines for 20 genes associated with T2D risk. We examined the impacts of each knockout on β cell differentiation, functions, and survival. We generated gene expression and chromatin accessibility profiles on β cells derived from each knockout line. Analyses of T2D-association signals overlapping HNF4A-dependent ATAC peaks identified a likely causal variant at the FAIM2 T2D-association signal. Additionally, the integrative association analyses identified four genes (CP, RNASE1, PCSK1N, and GSTA2) associated with insulin production, and two genes (TAGLN3 and DHRS2) associated with β cell sensitivity to lipotoxicity. Finally, we leveraged deep ATAC-seq read coverage to assess allele-specific imbalance at variants heterozygous in the parental line and identified a single likely functional variant at each of 23 T2D-association signals.

Functional interrogation of twenty type 2 diabetes-associated genes using isogenic hESC-derived β-like cells

Genetic studies have identified numerous loci associated with type 2 diabetes (T2D), but the functional role of many loci has remained unexplored. In this study, we engineered isogenic knockout human embryonic stem cell (hESC) lines for 20 genes associated with T2D risk. We systematically examined β-cell differentiation, insulin production and secretion, and survival. We performed RNA-seq and ATAC-seq on hESC-β cells from each knockout line. Analyses of T2D GWAS signals overlapping with HNF4A-dependent ATAC peaks identified a specific SNP as a likely causal variant. In addition, we performed integrative association analyses and identified four genes ( CP, RNASE1, PCSK1N and GSTA2 ) associated with insulin production, and two genes ( TAGLN3 and DHRS2 ) associated with sensitivity to lipotoxicity. Finally, we leveraged deep ATAC-seq read coverage to assess allele-specific imbalance at variants heterozygous in the parental hESC line, to identify a single likely functional variant at each of 23 T2D GWAS signals.

EPID-20. NOVEL GLIOBLASTOMA POPULATION-BASED HISTOLOGIC STAIN NORMALIZATION

Histopathologic evaluation has been an integral part of clinical diagnosis for central nervous system tumors, providing information essential for classification, management, and treatment of the disease. Hematoxylin and eosin (H&E) staining is routinely used in histology, providing detail of tissue morphology, structure, and cellular composition. MOTIVATION: Slide staining is rife with color intensity variations, mainly due to differences in materials and staining protocols among others. These variations introduce inaccuracies in downstream computational analysis and quantification of disease, disabling the generalization of computational models. To overcome these variations, current approaches arbitrarily select a slide within the cohort to normalize all slides of the cohort, leading to non-reproducible results in other cohorts. We develop a population-based whole slide image (WSI) normalization method based on overall region driven stain vectors and color histogram, weighted by corresponding percent contribution to overall slide (PCOS). METHODS: We identified a cohort of 509 H&E stained WSIs with corresponding anatomical annotations from the Ivy Glioblastoma Atlas Project. These WSIs and annotations were reviewed by two neuropathologists for correctly annotated regions. Each region was weighted according to PCOS, WSIs with PCOS < 0.05% were discarded. Then, the optical densities and histograms calculated. Resulting color histogram and optical density was applied to the WSI cohort. Finally, stain intensity variability pre- and post- normalization was compared. RESULTS: Normalizing WSIs based on our approach, results in a significant (p < 0.01, Wilcoxon) improvement in color intensity variation for eight of nine regions tested, with the exception of “Pseudopalisading Cells with no visible Necrosis” (p = 0.8). DISCUSSION: This novel transformative technique is insensitive to artificially staining background density and straightforward to apply. Furthermore, the approach shows promise towards a viable and robust tool for stain normalization in large WSIs cohorts, with the potential towards a stain normalization standard generalizable to other diseases.

NIMG-09. PREDICTING OVERALL SURVIVAL OF GLIOBLASTOMA PATIENTS ON MULTI-INSTITUTIONAL HISTOPATHOLOGY STAINED SLIDES USING DEEP LEARNING AND POPULATION-BASED NORMALIZATION

MOTIVATION

Glioblastoma is the most common and aggressive adult brain tumor. Clinical histopathologic evaluation is essential for tumor classification, which according to the World Health Organization is associated with prognostic information. Accurate prediction of patient overall survival (OS) from clinical routine baseline histopathology whole slide images (WSI) using advanced computational methods, while considering variations in the staining process, could contribute to clinical decision-making and patient management optimization.

METHODS

We utilize The Cancer Genome Atlas glioblastoma (TCGA-GBM) collection, comprising multi-institutional hematoxylin and eosin (H&E) stained frozen top-section WSI, genomic, and clinical data from 121 subjects. Data are randomly split into training (80%), validation (10%), and testing (10%) sets, while proportionally keeping the ratio of censored patients. We propose a novel deep learning algorithm to identify survival-discriminative histopathological patterns in a WSI, through feature maps, and quantitatively integrate them with gene expression and clinical data to predict patient OS. The concordance index (C-index) is used to quantify the predictive OS performance. Variations in slide staining are assessed through a novel population-based stain normalization approach, informed of glioblastoma distinct histologic sub-regions and their appearance from 509 H&E stained slides with corresponding anatomical annotations from the Ivy Glioblastoma Atlas Project (IvyGAP).

RESULTS

C-index was equal to 0.797, 0.713, and 0.703 for the training, validation, and testing data, respectively, prior to stain normalization. Following normalization, staining variations in H&E and ‘E’ gained significant improvements in IvyGAP (pWilcoxon< 0.01) and TCGA-GBM (pWilcoxon< 0.0001) data, respectively. These improvements contributed to further optimizing the C-index to 0.871, 0.777, and 0.780 for the training, validation, and testing data, respectively.

CONCLUSIONS

Appropriate normalization and integrative deep learning yield accurate OS prediction of glioblastoma patients through H&E slides, generalizable in multi-institutional data, potentially contributing to patient stratification in clinical trials. Our computationally-identified survival-discriminative histopathological patterns can contribute in further understanding glioblastoma.