Wavelet-based cancer drug recommender system

Abstract

Molecular nature of cancer is the foundation of systematic studies of cancer genomes, providing exceptional insights and allowing treatments advancement in clinic. We combine techniques of image processing for feature enhancement and recommender systems for proposing a personalized ranking of cancer drugs. We use a database containing drug sensitivity data for more than 310.000 IC50, describing response of more than 300 anticancer drugs across 987 cancer cell lines. The system is implemented in Python (Google Colaboratory) and succeed to find best fitted drugs for cancer cell lines. After several preprocessing tasks, regarding drug sensitivity data, two experiments are performed. First experiment uses original DNA microarray images and the second one uses wavelet transforms to preprocess images. Our main goal is to assess the impact of using wavelet transformed DNA microarray images (versus original images) on the proposed framework. The experiments show that, by improving the search of cancer cell lines with similar profile to the new cell line, wavelet transformed DNA microarray images produce better results, not only in terms of evaluation metrics (hit-rate and average reciprocal hit-rate), but also regarding execution time.