Fertility-GRU

Fertility-GRU predicts functions of fertility-related proteins by integrating Gated Recurrent Units (GRUs) with Position-Specific Scoring Matrix (PSSM) profiles to model sequential and evolutionary information in protein sequences.

Key Features:

• Gated Recurrent Units (GRUs): Employs GRUs to capture sequential dependencies in protein sequences.
• Position-Specific Scoring Matrix (PSSM): Incorporates PSSM profiles to leverage evolutionary information embedded in protein sequences.
• Predictive accuracy: Achieves cross-validation accuracy of 85.8% and independent test accuracy of 91.1%.
• Overfitting mitigation: Implements dropout layers within the deep learning architecture to reduce overfitting.
• Performance metrics: Independent testing reports sensitivity 90.5%, specificity 91.7%, and Matthews correlation coefficient (MCC) of 0.82.
• Comparative performance: Outperforms existing state-of-the-art predictors on the same dataset.

Scientific Applications:

• Bone marrow: Functional annotation of fertility-related proteins implicated in bone marrow contexts.
• Peripheral blood: Functional prediction of fertility-related proteins present in peripheral blood.
• Postnatal mammalian ovaries: Analysis of fertility-related protein functions in postnatal mammalian ovaries.
• Sperm production parameters: Investigation of proteins influencing parameters such as daily sperm production.

Methodology:

Training of deep neural networks using Gated Recurrent Units (GRUs) with Position-Specific Scoring Matrix (PSSM) profiles, including dropout layers to mitigate overfitting.