Deep neural networks, and combinations of supervised, semi-supervised and unsupervised learning techniques, are increasingly being used to solve difficult problems in medical image analysis, medical diagnosis, and analysis of biomarkers.
Deep learning is a branch of machine learning and often refers to artificial neural networks (“ANN”) that are composed of many layers. Broadly speaking, the goal of deep learning is to model complex, hierarchical features in data. It involves feeding a computer system a lot of data, which it can use to make decisions about other previously unseen data. Deep learning is about learning to predict in ways which can involve more complex dependencies between the input features.
In layman terms, neural networks are essentially computational models that attempt to mimic the action potential cascades that typify biological data processing (i.e. like our brain). So artificial neurons essentially act as little processors. ANNs are generally presented as systems of interconnected "neurons" which exchange messages between each other. These networks ask a series of binary true/false questions, or extract a numerical value, of every bit of data which pass through them, and classify it according to the answers received. The connections have numeric weights that can be tuned based on experience, making neural nets adaptive to inputs and capable of both supervised and unsupervised learning.
Thirty or forty years ago, neural networks were only two or three layers deep as it was not computationally feasible to build larger networks. Today it is common to have neural networks with 10+ layers and even 100+ layer ANNs are not unheard of. In fact, our DeepNEU models currently have thousands of layers.
Enter 123Genetix and DeepNEU
DeepNEU is our big data solution specifically engineered to empower rare diseases researchers. This highly novel deep learning computational platform combines (1) extensive and publicly available rare diseases gene network data with (2) modified Neutrosophic Logic, and (3) features of deep ANNs and Support Vector Machines.
The current version (v2.1) of DeepNEU continues to undergo testing and has been used to generate beta versions of several rare disease models for which the underlying genetics and clinical features are reasonably well characterized. A few examples of such well characterized diseases are Cystic Fibrosis, Duchenne Muscular Dystrophy and Sickle Cell Disease. These system biology models are easily customized to meet individual researchers’ needs. The purpose of these advanced models is to empower rare disease researchers to: (1) develop new disease hypotheses; (2) design and run experimental simulations to test those hypotheses; (3) write better grants and (4) identify potential disease specific biomarkers.
While DeepNEU itself is not a drug discovery engine, the identification of novel biomarkers should drive future research towards new and effective therapies where none exist now.