Cluster Identification in Metagenomics – A Novel Technique of Dimensionality Reduction through Autoencoders

Analysis of metagenomic data is not only challenging because they are acquired from a sample in their natural habitats but also the high volume and dimensionality. The fact that no prior lab based cultivation carried out metagenomics makes inference on presence numerous microorganisms all more challenging, accentuating need for an informative visualization this data. In successful visualization, congruent reads sequences should appear clusters depending diversity taxonomy sequenced sample. represented by oligonucleotide frequency vectors inherently dimensional therefore impossible to visualize as is. This raises dimensionality reduction technique convert these higher sequence into lower purposes. process, preservation genomic characteristics must be given highest priority. Currently, purposes metagenomics, Principal Component (PCA) which linear t-distributed Stochastic Neighbor Embedding (t-SNE), non-linear technique, widely used. Albeit wide use, techniques exceptionally suited domain with certain shortcomings weaknesses. Our research explores possibility using autoencoders, deep learning has potential overcome prevailing impediments existing eventually leading richer visualizations.