NDAR, a Resource to Help Define and Improve Phenotype and Sub-Phenotype Definition in Autism Research

Background:  :  The National Database for Autism Research (NDAR) supports data sharing for a broad array of clinical, genomic, and brain imaging autism research data.    

Objectives:  Categorizing research subjects by phenotype/sub-phenotype and even subtype is essential for those interested in exploring this rich resource.   Such categorizing is subject to debate, but by providing query access to tens of thousands of available assessments and data structures on research subjects, the community is able to critique, corroborate and further define the rules associated with the categorization. 

Methods:  Using data from the 100 projects now contributing data, we have begun a novel approach in addressing the fundamental need to categorize research subjects.  By using thousands of subjects across all of these labs, we have defined cutoffs that are supported initially by literature review, but then optimized by statistical analysis.  Such results may bias the categories but are normalized to the type of research currently being conducted. For the broader autism phenotype, we chose to sort research participants into three categories (mildly affected, affected, severely affected) only when ADOS and ADI were available along with scores from Vineland, SCQ, or an IQ measure. A subject is only assigned to the category if they met criteria for each assessment.  By using an automated rules pipeline, it is possible to rerun the categorization, optimizing the rules, until all subjects with a well characterized phenotype are defined

Results:  The NDAR team will present at IMFAR 2012 the rules supporting currently defined phenotypes/subphenotype of “Autism-Like Development Disorders” (the Rules for Severely Affected are provided below as an example) and the Minimally Verbal subtype.  Others may be defined by IMFAR 2012.  Furthermore, we will update IMFAR attendees on the progress of utilizing the same pipeline to aggregate across federated resources such as AGRE and SFARI.   

ADI-R

Total for Section A: Qualitative Abnormalities in Reciprocal Social Interaction > 10 AND

Total Section B:        Non-Verbal: Qualitative Abnormalities in Communication >7 OR

Total Section B:        Verbal: Qualitative Abnormalities in Communication > 8 AND

Total Section C:        Restricted, Repetitive, and Stereotyped Patterns of Behavior > 3 AND

Total Section D ≥1

ADOS                                            Module 1: Module 2: Module 3: Module 4:

Communication Total:                    > 4                > 5                > 3         > 3                 AND

Social Interaction Total:                  > 7                > 6                > 6         > 6                AND

Communication + Social Total:      > 12              > 12              >10        > 10

And at least one of these tests scores:

SCQ   

SCQ total score > 15 OR

IQ        

Less than average IQ < 85 OR

Vineland Survey

Less than average:

Composite domain total score < 85 AND

Communication domain total score < 85 AND

Living skills domain total score < 85 AND

Motor skills domain total score < 85 AND

Socialization domain total score < 85

Conclusions:  Defining and making categories (phenotypes/sub-phenotypes and subtype) available to the research community for data exploration on such a large scale is novel.    When such data is further refined and combined with other types of data from genomics studies and structural imaging, opportunities will be opened up for accelerated scientific discovery.