Bioinformatics and Scientific Programming Core Facility
- Ryan Dale, PhD, Scientific Information Officer
- Caroline Esnault, PhD, Bioinformatics Scientist
- Gennady Margolin, PhD, Bioinformatics Scientist
- Apratim Mitra, PhD, Bioinformatics Scientist
- Hongen (Henry) Zhang, PhD, Staff Scientist
- Sydney Hertafeld, BS, Postbaccalaureate Fellow
- Eva Jason, BS, Postbaccalaureate Fellow
- Nicholas Johnson, BS, Postbaccalaureate Fellow
- Arjun Mittal, BS, Postbaccalaureate Fellow
The goal of the Bioinformatics and Scientific Programming Core (BSPC) is to provide expert bioinformatics support to NICHD researchers, assisting at all stages from experimental design through multiple iterations of analysis to final manuscript preparation. In addition, we develop software tools that can be applied to a wide range of bioinformatics, genomics, and general data analysis, both at NICHD and in the larger international scientific community. We coordinate training for staff and trainees in basic programming and genomic analyses to help build bioinformatics support directly within labs.
Structure
The BSPC uses a “hub and spoke” model, consisting of a central core of staff (currently in Building 10) coordinating with embedded bioinformaticians (currently in Buildings 6 and 49) working directly in laboratories. This allows us to build centralized infrastructure that can be re-used across many research programs, while at the same time maintaining focused and custom local support in labs. Joint meetings and discussion allow everyone, central and embedded, to share lessons learned and identify new tools and methods.
Projects overview
In 2020, the BSPC worked on 65 projects, collaborating with PIs, fellows, staff scientists, and staff clinicians across 29 laboratories. Of these, 30 were new projects and 35 were carried over from the previous year. The projects included assays such as bulk RNA-seq, single-cell RNA-seq, ChIP-seq, whole-exome sequencing, whole-genome sequencing, DNA methylation, CUT&RUN, bulk ATAC-seq, and single-cell ATAC seq. New projects this year also included assessment of protein motifs, identification of co-evolving protein domains, SLAM-seq, and structural variant calling. Some projects involved custom algorithm development and tool development, and many projects required integration with published studies. Two new projects converted legacy code used in labs into a modern software architecture, which the BSPC can maintain into the future. Roughly a third of the projects involved RNA-seq differential expression analysis.
Projects often begin with an in-depth discussion with researchers to understand the background and goals of the project. It is important for us to understand the underlying biology and details of the experimental design (when applicable) for each project, so that we can make the most informed analysis decisions. We then provide a prioritized plan for the first round of analysis and schedule the work. There are often several iterations of analysis as a project progresses. Each iteration may add more sophisticated analyses, new data generated by the lab, or integrate results with published data. As expected for a no-cost shared resource, the time it takes for one iteration on one project is highly dependent on the existing workload across all other projects that we are handling in the Institute.
After each iteration, we meet to discuss the results in detail. The meeting includes a walk-through of the results, the computational background, discussion of how to use and interpret the tables, figures, and other output, and recommendations for next steps. Depending on the researchers’ interests, this can also include a discussion of the code and help with running it or adapting it to other projects in the lab. The next iteration of analysis is then planned, prioritized, and scheduled.
The BSPC’s collaboration includes writing the manuscript, producing figures and tables, consulting on interpretation, writing detailed computational methods, submitting data to public repositories, reviewing code, and submitting code to public repositories along with the complete software environments required to make the analyses reproducible.
Projects: computation and code
Most projects are multi-week or multi-month projects, which continue after many iterations and often require authoring substantial amounts of custom R and Python code. We work closely with NICHD's Molecular Genomics Core, where much of the raw high-throughput sequencing data for NICHD are generated. We can access these data directly, avoiding the need to coordinate data transfer and/or storage space with researchers. Analysis performed by the BSPC makes extensive use of NIH’s Biowulf high-performance computing cluster, and there is no direct cost to researchers for work done by the BSPC.
To ensure long-term computational reproducibility, we build a complete software environment for each project, which allows us to track all versions of software and dependencies, and any one project’s environment can be updated without affecting any others. All source code is kept under version control so that the entire history of the project can be tracked. We also build reproducible workflows for each project that keep track of which results have been updated and, wherever possible, provide output as standalone, interactive HTML files, so that researchers can easily explore their results.
Additional software development and computational resources
The BSPC continues to develop and maintain publicly available open-source tools. One example is lcdb-wf, a system of workflows and pipelines to process high-throughput sequencing data, run extensive quality control, and perform differential ChIP-seq or RNA-seq analyses and which runs on NIH’s Biowulf computing cluster. We also continue to contribute to the Bioconda project, a system used by bioinformaticians worldwide to easily install biology-related software tools.
The BSPC maintains an RStudio Connect Server instance, which allows us to publish interactive applications that researchers can use to interactively explore and plot their data. We also maintain a GitLab instance in NICHD’s data center, which provides source code version control, issue tracking, and documentation for projects we work on in such a way that they can be shared with collaborators. These repositories currently store tens of thousands of lines of Python and R code and documentation written by the BSPC and used in various projects.
Publications
- Chang E, Fu C, Coon SL, Alon S, Bozinoski M, Breymaier M, Bustos DM, Clokie SJ, Gothilf Y, Esnault C, Michael Iuvone P, Mason CE, Ochocinska MJ, Tovin A, Wang C, Xu P, Zhu J, Dale R, Klein DC. Resource: A multi-species multi-timepoint transcriptome database and webpage for the pineal gland and retina. J Pineal Res 2020;69(3):e12673.
- Kang H, Jha S, Ivovic A, Fratzl-Zelman N, Deng Z, Mitra A, Cabral WA, Hanson EP, Lange E, Cowen EW, Katz J, Roschger P, Klaushofer K, Dale RK, Siegel RM, Bhattacharyya T, Marini JC. Somatic SMAD3-activating mutations cause melorheostosis by up-regulating the TGF-ß/SMAD pathway. J Exp Med 2020;217:e20191499.
- Li L, Mitra A, Cui K, Zhao B, Choi S, Lee JY, Stamos DB, El-Khoury D, Warzecha C, Pfeifer K, Hardwick J, Zhao K, Venters B, Davé UP, Love PE. Ldb1 is required for Lmo2 oncogene-induced thymocyte self-renewal and T-cell acute lymphoblastic leukemia. Blood 2020;135:2252-2265.
- Melamed S, Adams PP, Zhang A, Zhang H, Storz G. RNA-RNA interactomes of ProQ and Hfq reveal overlapping and competing roles. Mol Cell 2019;77:411-425.
- Lee SY, Hung S, Esnault C, Pathak R, Johnson KR, Bankole O, Yamashita A, Zhang H, Levin HL. Dense transposon integration reveals essential cleavage and polyadenylation factors promote heterochromatin formation. Cell Rep 2020;30:2686-2698.
Collaborators
- William J. Pavan, PhD, Genomics, Development and Disease Section, NHGRI, Bethesda, MD
- Michael E. Ward, MD, PhD, Inherited Neurodegenerative Diseases Unit, NINDS, Bethesda, MD
Contact
For more information, email ryan.dale@nih.gov or visit https://www.nichd.nih.gov/about/org/dir/other-facilities/cores/bioinformatics.
