Platform Development, Complex Modeling (Core E)

As research becomes more interdisciplinary, there is an increasing need to combine many sources of data, such as genetic, neuroimaging, and behavioral metrics. Each approach generates data of multiple forms. While multimodal data are extremely rich and can be powerful windows for discovery, such data also present several challenges: (1) organizing data efficiently; (2) ensuring that data are put in a format ready for analysis (e.g., test scoring, condensing E-prime output to consolidated scores); and (3) analyzing data in a way that truly captures all the richness of multiple measures. Core E helps investigators develop platforms that are flexible, semi-automated, combine many types of data, and allow for rich queries.

  • Platform Development

    REDCap Integration. Core E, in collaboration with Cores B and C, provides services and software to automate behavioral and neuroimaging data management using REDCap databases as an integration platform. Such an approach allows for combining data from various modalities in a more seamless manner. Core C, in collaboration with Core E, has pioneered the use of REDCap’s Application Programming Interface (API) to implement large data export and import routines. Examples include:

    • Linking iPad/tablet data entry to REDCap
    • Providing automated routines for analyzing raw data in files and uploading results into REDCap
    • Providing automated routines to analyze MRI data
    • Developing automated routines to analyze mouse behavior
  • Complex Modeling

    Core E will spearhead developing and implementing new methods in IDD research using exemplars for IDDRC investigators. These methods will be presented at Core E seminars so that IDDRC investigators can understand how to relate these newer methods to their project-specific needs. For example, predictive methods based on machine-learning are increasingly used in neuroimaging or genomic data, but have been utilized to a lesser extent with behavioral data, and even less often in IDD research. While many machine-learning approaches have the disadvantage of typically needing large sample sizes, they also can offer more subtle examination of differences between populations.

Key Personnel