An Artificial Neural Network Integrated Pipeline for Biomarker Discovery Using Alzheimer's Disease as a Case Study

Abstract

The field of machine learning has allowed researchers to generate and analyse vast amounts of data using a wide variety of methodologies. Artificial Neural Networks (ANN) are some of the most commonly used statistical models and have been successful in biomarker discovery studies in multiple disease types. This review seeks to explore and evaluate an integrated ANN pipeline for biomarker discovery and validation in Alzheimer's disease, the most common form of dementia worldwide with no proven cause and no available cure. The proposed pipeline consists of analysing public data with a categorical and continuous stepwise algorithm and further examination through network inference to predict gene interactions. This methodology can reliably generate novel markers and further examine known ones and can be used to guide future research in Alzheimer's disease.

Keywords: AD, Alzheimer's disease; ANN, artificial neural network; APP, amyloid precursor protein; Alzheimer's disease; Artificial neural network; Aβ, beta amyloid; Biomarker discovery; MLP, multi-layer perceptron; Machine learning; NFT, neurofibrillary tangles; Network inference; Supervised learning.

Fig. 1

Physiological differences between a healthy and AD brain section, demonstrating white matter shrinkage in the hippocampus and cerebral cortex.

Fig. 2

Amyloid plaques (pink) and neurofibrillary tangles (black) in Alzheimer's disease brain tissue.

Fig. 3

Diagram of the amyloid cascade hypothesis showing the theorised links between the aggregation of Aβ to cell death and dementia.

Fig. 4

Microglial cell diagram showing the formation of the NLRP3 inflammasome and cytokine cascade as a result of Aβ detection.

Fig. 5

Workflow diagram of the artificial neural network algorithm developed by Lancashire et al. [31] used for this project. The parameters for the hidden and output layer nodes are in their paper.

Fig. 6

Force directed interactome encompassing 500 gene probes and 1000 predicted interactions of the hippocampus in the E-GEOD-48350 AD cohort. Red edges indicate and inhibitory effect, whereas blue edges indicate promotion. Edge thickness is directly proportional to the strength of the interaction. Green nodes are upregulated genes while red ones are downregulated. The intensity of the colour is directly proportional to the degree of up- or downregulation.

Fig. 7

Alternative circular layout interactome of the 1000 strongest interactions between 500 genes in AD independent of the brain region in the E-GEOD-48350 dataset. Based on the overall expression of all brain regions. Novel targets identified. Red edges indicate and inhibitory effect, whereas blue edges indicate promotion. Edge thickness is directly proportional to the strength of the interaction. Green nodes are upregulated genes while red ones are downregulated. The intensity of the colour is directly proportional to the degree of up- or downregulation.

Fig. 8

Focused Tubulin interactome based on Fig. 7. Tubulin beta 2A interactions in AD. Of note is its positive regulation by an NFKB inhibitor.