Search

Patient-derived xenografts (PDXs) model human intra-tumoral heterogeneity in the context of the intact tissue of immunocompromised mice. Histological imaging via hematoxylin and eosin (H&E) staining is performed on PDX samples for routine assessment and, in principle, captures the complex interplay between tumor and stromal cells. Deep learning (DL)-based analysis of large human H&E image repositories has extracted inter-cellular and morphological signals correlated with disease phenotype and therapeutic response. Here, we present an extensive, pan-cancer repository of nearly 1,000 PDX and paired human progenitor H&E images. These images, curated from the PDXNet consortium, are associated with genomic and transcriptomic data, clinical metadata, pathological assessment of cell composition, and, in several cases, detailed pathological annotation of tumor, stroma, and necrotic regions. We demonstrate that DL can be applied to these images to classify tumor regions with an accuracy of 0.87. Further, we show that DL can predict xenograft-transplant lymphoproliferative disorder, the unintended outgrowth of human lymphocytes at the transplantation site, with an accuracy of 0.97. This repository enables PDX-specific investigations of cancer biology through histopathological analysis and contributes important model system data that expand on existing human histology repositories. We expect the PDXNet Image Repository to be valuable for controlled digital pathology analysis, both for the evaluation of technical issues such as stain normalization and for development of novel computational methods based on spatial behaviors within cancer tissues. Citation Format: Brian S. White, Xing Yi Woo, Soner Koc, Todd Sheridan, Steven B. Neuhauser, Shidan Wang, Yvonne A. Evrard, John David Landua, R Jay Mashl, Sherri R. Davies, Bingliang Fang, Maria Gabriela Raso, Kurt W. Evans, Matthew H. Bailey, Yeqing Chen, Min Xiao, Jill Rubinstein, Ali Foroughi pour, Lacey Elizabeth Dobrolecki, Maihi Fujita, Junya Fujimoto, Guanghua Xiao, Ryan C. Fields, Jacqueline L. Mudd, Xiaowei Xu, Melinda G. Hollingshead, Shahanawaz Jiwani, PDXNet consortium, Tiffany A. Wallace, Jeffrey A. Moscow, James H. Doroshow, Nicholas Mitsiades, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis G. Carvajal-Carmona, Alana L. Welm, Bryan E. Welm, Ramaswamy Govindan, Shunqiang Li, Michael A. Davies, Jack A. Roth, Funda Meric-Bernstam, Yang Xie, Meenhard Herlyn, Li Ding, Michael T. Lewis, Carol J. Bolt, Dennis A. Dean, Jeffrey H. Chuang. A pan-cancer PDX histology image repository with genomic and pathological annotations for deep learning analysis. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5407.

We created the PDX Network (PDXNet) portal (https://portal.pdxnetwork.org/) to centralize access to the National Cancer Institute-funded PDXNet consortium resources, to facilitate collaboration among researchers and to make these data easily available for research. The portal includes sections for resources, analysis results, metrics for PDXNet activities, data processing protocols and training materials for processing PDX data. Currently, the portal contains PDXNet model information and data resources from 334 new models across 33 cancer types. Tissue samples of these models were deposited in the NCI’s Patient-Derived Model Repository (PDMR) for public access. These models have 2134 associated sequencing files from 873 samples across 308 patients, which are hosted on the Cancer Genomics Cloud powered by Seven Bridges and the NCI Cancer Data Service for long-term storage and access with dbGaP permissions. The portal includes results from freely available, robust, validated and standardized analysis workflows on PDXNet sequencing files and PDMR data (3857 samples from 629 patients across 85 disease types). The PDXNet portal is continuously updated with new data and is of significant utility to the cancer research community as it provides a centralized location for PDXNet resources, which support multi-agent treatment studies, determination of sensitivity and resistance mechanisms, and preclinical trials.

We created the PDX Network (PDXNet) Portal (https://portal.pdxnetwork.org/) to centralize access to the National Cancer Institute-funded PDXNet consortium resources (i.e., PDX models, sequencing data, treatment response data, and bioinformatics workflows), to facilitate collaboration among researchers, and to make resources easily available for research. The portal includes sections for resources, analysis results, metrics for PDXNet activities, data processing protocols, and training materials for processing PDX data.The initial portal release highlights PDXNet model and data resources, including 334 new models across 33 cancer types. Tissue samples of these models were deposited in the NCI’s Patient-Derived Model Repository (PDMR) for public access. These models have 2,822 associated sequencing files from 873 samples across 307 patients, which are hosted on the Cancer Genomics Cloud powered by Seven Bridges and the NCI Cancer Data Service for long-term storage and access with dbGaP permissions. The portal also includes results from standardized analysis workflows on PDXNet sequencing files and PDMR data (2,594 samples from 463 patients across 78 disease types). These 15 analysis workflows for whole-exome and RNA-Seq data are freely available, robust, validated, and standardized.The model and data lists will grow substantially over the next two years and will be continuously updated as new data are available. PDXNet models support multi-agent treatment studies, determination of sensitivity and resistance mechanisms, and preclinical trials. The PDXNet portal is a centralized location for these data and resources, which we expect to be of significant utility for the cancer research community.

Developments in high-throughput sequencing (HTS) result in an exponential increase in the amount of data generated by sequencing experiments, an increase in the complexity of bioinformatics analysis reporting and an increase in the types of data generated. These increases in volume, diversity and complexity of the data generated and their analysis expose the necessity of a structured and standardized reporting template. BioCompute Objects (BCOs) provide the requisite support for communication of HTS data analysis that includes support for workflow, as well as data, curation, accessibility and reproducibility of communication. BCOs standardize how researchers report provenance and the established verification and validation protocols used in workflows while also being robust enough to convey content integration or curation in knowledge bases. BCOs that encapsulate tools, platforms, datasets and workflows are FAIR (findable, accessible, interoperable and reusable) compliant. Providing operational workflow and data information facilitates interoperability between platforms and incorporation of future dataset within an HTS analysis for use within industrial, academic and regulatory settings. Cloud-based platforms, including High-performance Integrated Virtual Environment (HIVE), Cancer Genomics Cloud (CGC) and Galaxy, support BCO generation for users. Given the 100K+ userbase between these platforms, BioCompute can be leveraged for workflow documentation. In this paper, we report the availability of platform-dependent and platform-independent BCO tools: HIVE BCO App, CGC BCO App, Galaxy BCO API Extension and BCO Portal. Community engagement was utilized to evaluate tool efficacy. We demonstrate that these tools further advance BCO creation from text editing approaches used in earlier releases of the standard. Moreover, we demonstrate that integrating BCO generation within existing analysis platforms greatly streamlines BCO creation while capturing granular workflow details. We also demonstrate that the BCO tools described in the paper provide an approach to solve the long-standing challenge of standardizing workflow descriptions that are both human and machine readable while accommodating manual and automated curation with evidence tagging. Database URL: https://www.biocomputeobject.org/resources

The Cancer Genomics Cloud (CGC), powered by Seven Bridges, is an NCI-funded platform that streamlines access to large cancer datasets, bioinformatic tools, and cloud computation for cancer researchers. The attributes of the CGC designed to democratize data analysis also make it ideal for training the next generation of data scientists. During the Covid-19 pandemic, it has become clear that remote/virtual learning is of great importance for workforce development, and that reducing barriers to high quality educational resources is critical for many populations. Here we present our best practices for education of bioinformatics using the CGC, taking advantage of both distributed cloud networks and platform features to enhance learning. Our best practices methods are focused on a systematic approach that takes the instructor and students through a typical bioinformatics workflow in their field of research. Briefly, the organizational structure of the CGC, known as “projects,'' contains all aspects of an analysis, including data files, tools, and tool settings. Projects have fine-grained permission settings, which allow the owner to securely share their project for viewing or editing. An instructor can generate an example analysis from start to finish then share an entire project with trainees. Both students and instructors have access to the same data on the cloud, so the teacher can pre-populate the projects with specific files, ensuring the same starting point. All members within the project can communicate, including adding notes directly on the results, allowing students to troubleshoot in real time with the teacher. Instructors can manage costs for the whole class through the CGC’s billing system. The CGC also provides Public Projects for self-guided training, where a researcher can learn by example using the data, tools, and completed tasks within the project, including detailed instructions embedded in markdown language. We have successfully used these methods to train masters students in bioinformatics at Georgetown University for three years, as well as high school students, college students, and current cancer researchers. This approach to virtual learning of bioinformatics can democratize training for the next generation of data scientists. The FAIR practices built into the CGC enables education across all levels of expertise, empowering users to drive cancer research from any stage of their career. Citation Format: Manisha Ray, Sai Lakshmi Subramanian, David Roberson, Zelia Worman, Dennis A. Dean, Soner Koc, Jack DiGiovanna, Brandi Davis-Dusenbery. The Cancer Genomics Cloud enables bioinformatics training in a hybrid educational world [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 6393.

Applying machine learning algorithms to analysis of multi-omics datasets has provided a wealth of analysis and interpretations of data not easily achievable by conventional methods. The Cancer Genomics Cloud (CGC), powered by Seven Bridges, offers many helpful features to simplify and streamline the performance of machine learning (ML) tasks on the platform. The CGC platform contains features and tools for all of the steps of a ML project, from data exploration, to model generation, to production, so that all key steps occur on the platform. The Data Cruncher interactive analysis tool enables users to perform data exploration, modeling, training, and visualization using familiar frameworks like Jupyterlab and RStudio. We applied this ML methodology to imaging tools, to create The SB Image Processing Toolkit. Here we demonstrate the use of this toolkit with specific cancer datasets, including training and analysis, with all steps taking place on the cloud. The SB Image Processing Toolkit is a collection of various deep-learning, preprocessing, and utility tools and workflows created with the purpose of performing image class prediction on any type of image data. The toolkit supports a variety of images used in cancer research, such as histopathology and radiology images, and common formats like JPG or PNG. There are nine distinct tools within the toolkit, including methods for quality control, normalization, preprocessing for histopathology or radiology images, classification, and deep learning. Together, these tools create an easy-to-use infrastructure to enable various stages of ML and Image Processing. Most importantly, it was designed to enable users to create and use ML image classifiers on the CGC without any coding experience. All workflows can be searched for and implemented via the Public Apps Gallery on the CGC. Users can import image data, run analysis, and compare results all in one environment. By harnessing the scale and flexibility of cloud computing, the SB Image Processing Toolkit brings massive speed and price improvements compared to manual writing and running model training scripts. In conjunction with the access to data and ease of use, the CGC makes complex machine learning methods accessible to any researcher. Citation Format: Soner Koc, Jovana Babic, Nevena Nikolic, Ana Stankovic, Daniel Ventre, Manisha Ray, Zelia F. Worman, Sai Lakshmi Subramanian, Dennis A. Dean, Jack DiGiovanna, Brandi Davis Dusenbery. The SB Image Processing Toolkit: Machine learning for cancer research on the CGC [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 6394.

BackgroundThe field of bioinformatics has grown at such a rapid pace that a gap in standardization exists when reporting an analysis. In response, the BioCompute project was created to standardize the type and method of information communicated when describing a bioinformatic analysis. Once the project became established, its goals shifted to broadening awareness and usage of BioCompute, and soliciting feedback from a larger audience. To address these goals, the BioCompute project collaborated with precisionFDA on a crowdsourced challenge that ran from May 2019 to October 2019. This challenge had a beginner track where participants submitted BCOs based on a pipeline of their choosing, and an advanced track where participants submitted applications supporting the creation of a BCO and verification of BCO conformance to specifications.ResultsIn total, there were 28 submissions to the beginner track (including submissions from a bioinformatics master’s class at George Washington University) and three submissions to the advanced track. Three top performers were selected from the beginner track, while a single top performer was selected for the advanced track. In the beginner track, top performers differentiated themselves by submitting BCOs that included more than the minimally compliant content. Advanced track submissions were very impressive. They included a complete web application, a command line tool that produced a static result, and a dockerized container that automatically created the BCO as the tool was run. The ability to harmonize the correct function, a simple user experience, and the aesthetics of the tool interface differentiated the tools.ConclusionsDespite being new to the concept, most beginner track scores were high, indicating that most users understood the fundamental concepts of the BCO specification. Novice bioinformatics students were an ideal cohort for this Challenge because of their lack of familiarity with BioCompute, broad diversity of research interests, and motivation to submit high-quality work. This challenge was successful in introducing the BCO to a wider audience, obtaining feedback from that audience, and resulting in a tool novices may use for BCO creation and conformance. In addition, the BCO specification itself was improved based on feedback illustrating the utility of a “wisdom of the crowd” approach to standards development.

We created the PDX Network (PDXNet) Portal to provide an intuitive way for researchers to explore and understand the models, sequencing data, and bioinformatics workflows generated by NCI's PDXNet consortium for research access (https://portal.pdxnetwork.org/). The portal also provides metrics for PDXNet's activities, data processing protocols, and training materials for processing PDX data. The PDXNet Portal highlights model and data resources that include 216 new models across 29 cancer types. The most prevalent cancers represented in the PDX model dataset include invasive breast carcinoma (30.6%), melanoma (18.1%), and adenocarcinoma (14.4%). PDXNet teams have provided 2263 sequencing files from 356 samples across 204 patients, comprising whole exome (82.9%) and RNA seq files (17.1%). The most prevalent cancers represented in the PDXNet sequencing data set include Breast Pleural Effusion (27.2%), Breast Poorly Differentiated (12.5%), and Lung Adenocarcinoma (9.6%). The portal also provides access to 9492 sequencing files across 78 disease types that include 2594 samples across 463 patients uploaded from the NCI Patient-Derived Model Repository. The dataset includes both whole exomes (52.8%) and RNA seq (47.2%) data. The PDMR samples include PDX (82.7%), primary tumor (5.7%), normal germline (5.5), organoid culture (3.2), and Mixed Tumor Culture (2.9). The PDMR dataset also has multiple passages: P0 (21.8%), P1(39.5%), P2 (25.6%), and P3 (8.5%). These models and data resources support ten PDXNet Pilot activities, multiple publications, and international collaborations. PDXNet has also developed a set of 13 robust, validated, and standardized workflows for processing PDXNet whole-exome and RNA seq data. Collectively, these workflows allow for the standardized processing of PDX and complementary human tissues (normal and tumor). Our plan is to continuously update the model and data lists on the PDX portal as resources are generated. We expect that the PDXNet generated models, scheduled to grow to 1000 new models by 2022, will support multi-agent treatment studies, determination of mechanisms of sensitivity and resistance, and pre-clinical trials for example through the COMBO-MATCH program. The robust standard workflows currently processing all PDX sequencing data may also facilitate harmonizing data across studies. Lastly, we expect that the generated sequencing data will support computational approaches for studying cancer evolution and the mechanisms underlying cancer treatments. Citation Format: Soner Koc, Mike Lloyd, Steven Neuhauser, Javad Noodbakhsh, Anuj Srivastava, Xing Yi Woo, Ryan Jeon, Jeffrey Grover, Sara Seepo, Christian Frech, Jack DiGiovanna, PDXNet Consortium, Yvonne A. Evard, Tiffany Wallace, Jeffrey Moscow, James H. Doroshow, Nicholas Mitsuade, Salma Kaochar, Chong-xian Pan, Moon S. Chen, Luis Carvarjal-Carmona, Alana Welm, Bryan Welm, Michael T. Lewis, Govindan Ramaswamy, Li Ding, Shunquang Li, Meenherd Herlyn, Mike Davies, Jack Roth, Funda Meric-Bernstam, Peter Robinson, Carol J. Bult, Brandi Davis-Dusenbery, Dennis A. Dean, Jeffrey H. Chuang. Advancing PDX research through model, data, and bioinformatics with the PDXNet Portal [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 3017.