Empowering protein expression images based diagnostics for breast cancer pharmacogenomics using deep learning

Abstract

The study aims at developing a patient specific pharmacogenomic mutation detection system for triple negative breast
cancer (TNBC) using the images of protein expression analysis by making use of deep learning technique. The images of
eight mutations related to TNBC have been considered such as AGT, CCL4, ESR1, IKBKB, MELK, NFKB, PDGF and
TP53. The pharmacogenomic analysis of these genes were carried out consisting of genomic, epigenomic, metagenomic
and environmental genomic analysis. The genetic signatures behind these factors were also identified along with the
specific protein expression images. The deep learning approach used in the work for the mutation identification makes use
of „one-short learning algorithm‟ which has given a prediction accuracy of about 81.75%. A secure sharing of patient data
has been set up through cryptographic methods as well as block chain technology. Hence this model seems to be an
ingenious, economical and easy way of identifying the mutations behind TNBC.