• Team Lead:
  • Evan Savage, CBT
• Team Members:
  • Daniel Darnell, Hartwell Center
  • Dale Hedges, Center for Applied Bioinformatics
  • Susanna Downing, Center for Pediatric Neurological Disease Research

Team 1 developed the Hackathon Evaluation Analysis for Recombination Deficiency (HEARD) Shiny app. This app provides an easy-to-use for interactive exploration of homologous recombination deficiency data at both the cohort and individual levels. It can generate a PDF report with important metrics, including the HRD Sum score as well as the COSMIC mutational signature exposure for HRD positive samples. 

• Team Lead: 
  • Christy LaFlamme, St. Jude Graduate Student
    
• Team Members:
  
  • Nadhir Djekidel, Center for Applied Bioinformatics
    
  • Pandurang Kolekar, Computational Biology
    
  • Wojciech Rosikiewicz, Center for Applied Bioinformatics

DNA methylation is an epigenetic modification that regulates gene expression, the dysregulation of which can cause various diseases. DNA methylation generally occurs at CpG sites, and >850,000 of these CpG sites can be readily assessed from genome-wide, cost-effective methylation arrays. However, few pipelines exist for analyzing and visualizing methylation array data, particularly with an emphasis on rare, disease-causing differentially methylated regions (DMRs). MethMiner is a convenient pipeline to process, analyze, and visualize methylation array data. MethMiner takes raw iDAT files as input, performs quality control/normalization, and uses a window analysis script to perform outlier DMR analysis on autosomes and sex chromosomes. DMRs are then annotated to inform functional interpretation. MethMiner also has data visualization capabilities through JupyterDash where the user can interact with their data through dimensionality reduction plots, DMR annotation breakdowns, and a genomic track browser. Ultimately, MethMiner seeks to provide an easily accessible and convenient methylation array pipeline, thereby expediting DNA methylation discoveries in disease research that can be translated into the clinic.

• Team Lead:
  • Wenan Chen, Center for Applied Bioinformatics
    
• Team Members:
  • Hyunjin Kim
    
  • Wenjian Yang, Pharmaceutical Sciences
    
  • Cody Ramirez, Developmental Neurobiology
    
  • Wenchao Zhang, Center for Applied Bioinformatics

Pathogenic mutations/variants are generally considered rare, however some variants, though rare in a global population, are enriched in specific backgrounds. Identification of a sample’s population or ethnicity can help resolve potential false positives and prioritize rare variants. Team 3 worked on a user-friendly R Shiny app to aid in this classification for individual samples and visualize the user’s input sample with existing samples from the 1000 Genomes Project.

• Team Lead: 
  • Stanley Pounds, Biostatistics
• Team Members:
  • Tushar Patni, Biostatistics
  • Yimei Li, Biostatistics
  • Subodh Selukar, Biostatistics

Cox proportional hazard model is a very common statistical model, usually presented as a table. Team 4 built a lightweight Shiny app called by a single command to interact with a cox model to interrogate it further with a thought towards personalized medicine. It allows users to visualize and compare models as well as interactively adjust parameters to look at goodness of fit in near real time.

• Team Lead:
  
  • Jared Andrews, Developmental Neurobiology
• Team Members:
  
  • Lawryn Kasper, Developmental Neurobiology
    
  • Jacob Steele, Center for Advanced Genome Engineering

Peak calling is a common, and often frustrating, task for epigenomic data due to varying signal to background ratios, heterogeneous peak profiles, and inconsistent antibody efficiencies resulting in seemingly endless rounds of parameter optimization. To address these challenges, Team 5 developed peakPeakeR, a R Shiny app allows users to easily compare the output of peak calling parameters from several tools on their data in near real time. Notably, peakPeekeR handles all peak caller installation and running one or more instances of each peak caller seamlessly. User data and each tools’ peak calls are displayed in a genome browser within the app, which currently supports MACS, MACS2, genrich, and Sicer2.

• Team Lead:
  • Jaimin Patel, Structural Biology
    
• Team Members:
  
  • Kubra Bahcivanci, Hematology
    
  • Pei-Lin Chen, Hematology
    
  • Sujuan Jia, Pathology
    
  • Karthik Vadambacheri Manian, Information Services
    
  • Kennon Silence, Facilities Operation and Maintenance

Tracking sample metadata and data locations is a major struggle for many groups and labs. Team Six designed SimplySamples utilizing relational databases to build a flexible, yet robust system to address this need. Their aim was to build an interface that would be useable by those with limited knowledge of databases that are used to using tools such as Excel for data tracking. An intuitive web UI allows users to query these databases to find the information they need, providing increased data access and fidelity. In time, the team hopes to add the ability to provide streamlined reports in a consistent fashion.

• Team Leads:
  • Jared Becksfort
  • Austin Faught, Radiation Oncology
    
• Team Members:
  • John Grime, Information Services
    
  • Lydia Wilson, Radiation Oncology
    
  • Rico Zhang, Radiation Oncology
    
  • Estelle Batin, Radiation Oncology
    
  • Franz Parkins, Information Services
    
  • Amber Bryant, Radiation Oncology

Team 7 created a simulation for the St. Jude Proton Therapy Center to develop a 3D collision avoidance application. They first created a 3D model of the therapy machine, from which their application can identify if a chosen treatment geometry is allowable, uncertain, or not possible and provide simple text feedback. Their model also allows 3D navigation of a mock treatment room by moving the three key components of the treatment machine. They plan to continue to build out this model with an eventual goal of gamifying it so that patients can see how the proton therapy will work and ultimately be calmer and more prepared before treatment.

• Team Lead: 
  • Xun Zhu, Center for Applied Bioinformatics

Team 8 was able to create an extremely performant web-based data viewer and editor, capable of handling millions of records. Full-stack row virtualization and WebWorker-enabled parallel parsing enable real-time streaming of records with no latency or performance hit. This is a valuable component for more complex web applications that deal with large databases or real-time data streaming.

• Team Lead:
  • Mia Panlilio, Center for Bioimage Informatics
    
• Team Members:
  
  • Jennifer Ocasio Adorno, Developmental Neurobiology
    
  • Anna Medyukhina, Center for Bioimage Informatics
    
  • Justin Williams, Tumor Cell Biology
    
  • Liam Hallada, St. Jude Graduate Student
    
  • Burkhard Hoeckendorf, Center for Bioimage Informatics
    
  • Sukesh Sangam, Center for Bioimage Informatics

Team 9 focused on optimizing tools for identifying and classifying cells with cytoplasmic markers, focusing specifically on brain microglia. They were able to set ground truths for cells with unique shapes using manual segmentation, then trained CellPose on these manually segmented images. This drastically improved the accuracy of identifying microglia in dense tissue sections. They also were able to improve speed by performing object detection to identify regions of interest rather than processing the entire image. Once they conquered identifying the cell type of interest, they were able to use clustering to detect microglia in differing stages of activation.