June 2013
Volume 54, Issue 15
Free
ARVO Annual Meeting Abstract  |   June 2013
Positive and Unlabeled Learning for Prioritizing Candidate Variants in Retinal Degenerative Diseases
Author Affiliations & Notes
  • Alex Wagner
    Genetics, University of Iowa, Iowa City, IA
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Kyle Taylor
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Adam DeLuca
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Thomas Casavant
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Edwin Stone
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Robert Mullins
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Todd Scheetz
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Terry Braun
    Ophthalmology and Visual Sciences, University of Iowa, Iowa City, IA
  • Footnotes
    Commercial Relationships Alex Wagner, None; Kyle Taylor, None; Adam DeLuca, None; Thomas Casavant, None; Edwin Stone, None; Robert Mullins, Alcon Research Ltd (F); Todd Scheetz, None; Terry Braun, Alcon Research, LTD (F)
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2013, Vol.54, 3376. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Alex Wagner, Kyle Taylor, Adam DeLuca, Thomas Casavant, Edwin Stone, Robert Mullins, Todd Scheetz, Terry Braun; Positive and Unlabeled Learning for Prioritizing Candidate Variants in Retinal Degenerative Diseases. Invest. Ophthalmol. Vis. Sci. 2013;54(15):3376.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose: Exome sequencing experiments are useful in identifying genetic variants that are causative of rare diseases with extreme genetic heterogeneity, such as retinal degenerative diseases. However, the number of plausible disease-causing variations identified in such studies are often overwhelming, even after applying expert-guided filtering steps. Thus, we seek to prioritize lists of candidate variants identified in exome sequencing studies by their likelihood to contribute to a retinal degenerative disease phenotype.

Methods: We leveraged publicly available data to quantitatively describe all genes in the context of relevant features. Specifically, we selected microarray data of gene expression in 10 tissues of the human eye, RNA-seq data in human retinal tissue and 16 other body tissues, and ChIP-seq data profiling CRX binding sites across the genome. We developed a novel method (Positive and Unlabeled Learning for Prioritization—PULP) to rank candidate disease-causing variants, and trained it on these data.

Results: Using a Monte Carlo simulation, PULP was shown to perform significantly better than random gene ordering (p < 1 x 10-6). Thirteen published RP linkages, all with identified disease-causing genes, were used to evaluate our system. The causative gene was prioritized at the top of the list 62% of the time, and within the top 5 results 77% of the time. These results were highly significant (p = 6 x 10-6). In addition, we demonstrate that our algorithm outperforms ENDEAVOUR, a current state-of-the-art technique in this field, which had successfully prioritized an RP causative variant in a familial study on Usher’s Syndrome. Our system also successfully prioritized the RP-causing DHDDS variant as the #1 candidate from a list of 20 from a recent exome study, and prioritized a recently identified RP-causing variant in MAK to #4 in a list of 348 candidates.

Conclusions: The PULP retinal degenerative disease model represents a huge step forward in the identification of retinal degenerative disease variants from exome sequencing studies. Moreover, this technique represents an unbiased, generalizable approach to integrating disease-specific quantitative genetic features for the purpose of disease-associated rankings in candidate lists.

Keywords: 539 genetics • 473 computational modeling • 696 retinal degenerations: hereditary  
×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×