Gnocis

Gnocis provides genome-wide analysis and machine-learning modelling of cis-regulatory elements (CREs), enabling prediction and characterization of promoters, enhancers, Polycomb/Trithorax Response Elements (PREs), silencers, and insulators from DNA sequence feature sets.

Key Features:

• Extensible APIs: APIs for integrating custom DNA sequence feature sets with machine learning models.
• Feature set implementation: Includes motif pair occurrence frequencies and the k-spectrum mismatch kernel to capture sequence-level regulatory signals.
• Machine learning integration: Interoperates with Scikit-learn and TensorFlow for training and evaluation of predictive models.
• Data handling tools: Supports management of sequence, region, and curve data for DNA bioinformatics analyses.
• Deep-MOCCA architecture: Provides a neural network architecture inspired by SVM-MOCCA that aims to generalize without prior motif knowledge.
• Performance optimization: Implemented in Cython and compiled for Python 3 to improve execution performance.

Scientific Applications:

• Genome-wide CRE prediction: Prediction and genome-wide scanning for candidate promoters, enhancers, PREs, silencers, and insulators from sequence-derived features.
• Drosophila PRE modelling: Application to Drosophila melanogaster Polycomb/Trithorax Response Elements, including use of Convolutional Neural Networks (CNNs) to model PREs.

Methodology:

Combines DNA sequence feature sets with machine-learning methods including motif pair occurrence frequencies and the k-spectrum mismatch kernel; integrates with Scikit-learn and TensorFlow; implements the Deep-MOCCA neural architecture inspired by SVM-MOCCA; Convolutional Neural Networks have been applied to PRE modelling; code is implemented in Cython for Python 3.