Ovary transcriptome profiling via artificial intelligence reveals a transcriptomic fingerprint predicting egg quality in striped bass, Morone saxatilis

PLoS One. 2014 May 12;9(5):e96818. doi: 10.1371/journal.pone.0096818. eCollection 2014.


Inherited gene transcripts deposited in oocytes direct early embryonic development in all vertebrates, but transcript profiles indicative of embryo developmental competence have not previously been identified. We employed artificial intelligence to model profiles of maternal ovary gene expression and their relationship to egg quality, evaluated as production of viable mid-blastula stage embryos, in the striped bass (Morone saxatilis), a farmed species with serious egg quality problems. In models developed using artificial neural networks (ANNs) and supervised machine learning, collective changes in the expression of a limited suite of genes (233) representing <2% of the queried ovary transcriptome explained >90% of the eventual variance in embryo survival. Egg quality related to minor changes in gene expression (<0.2-fold), with most individual transcripts making a small contribution (<1%) to the overall prediction of egg quality. These findings indicate that the predictive power of the transcriptome as regards egg quality resides not in levels of individual genes, but rather in the collective, coordinated expression of a suite of transcripts constituting a transcriptomic "fingerprint". Correlation analyses of the corresponding candidate genes indicated that dysfunction of the ubiquitin-26S proteasome, COP9 signalosome, and subsequent control of the cell cycle engenders embryonic developmental incompetence. The affected gene networks are centrally involved in regulation of early development in all vertebrates, including humans. By assessing collective levels of the relevant ovarian transcripts via ANNs we were able, for the first time in any vertebrate, to accurately predict the subsequent embryo developmental potential of eggs from individual females. Our results show that the transcriptomic fingerprint evidencing developmental dysfunction is highly predictive of, and therefore likely to regulate, egg quality, a biologically complex trait crucial to reproductive fitness.

Publication types

  • Research Support, Non-U.S. Gov't
  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Animals
  • Artificial Intelligence
  • Bass / embryology
  • Bass / metabolism*
  • Female
  • Gene Expression Profiling / methods*
  • Neural Networks, Computer
  • Ovary / embryology
  • Ovary / metabolism*
  • Ovum / metabolism*
  • Transcriptome / genetics*

Grants and funding

This work was supported by core funding of the Center of Excellence in Oceans and Human Health Center for Marine Genomics at Hollings Marine Laboratory (http://www.carolinas.noaa.gov/spotlight/hollings_0609.html) where the microarray analyses were performed, special grants (2008-34368-191136 and 2009-34368-19776) from the U.S. Department of Agriculture (USDA) National Institute of Food and Agriculture (NIFA) (http://www.csrees.usda.gov/), grants (R/AF-49, R/12-SSS-3, and R/MG-1019) from North Carolina Sea Grant (http://www.ncseagrant.org/), grants (R41-1 and R/CF-19) from the SC Sea Grant Consortium (http://www.scseagrant.org/), and a special award from the USDA-NIFA, National Animal Genome Research Project (NRSP-8) - Aquaculture Genome Project (http://www.animalgenome.org/aquaculture/) for microarray costs. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.