Logo: to the web site of Uppsala University

Cancer is a disease that can be characterized by an uncontrolled division of cells and the formation of tumors. Globally, cancer is the leading cause of death. There are various standard treatments available for cancer. These include chemotherapy, radiotherapy, and surgery, of which the first option is the least invasive. The chemotherapeutic drugs have the drawback of possible toxicity to healthy cells and may cause drug resistance. However, a newer and better drug delivery option is now available. This is the Nanoparticle-based carrier, whose benefits include decreased drug resistance, enhanced specificity and improved pharmacokinetics of the anticancer drugs. Nanoparticles also have several other applications like imaging, radiotherapy, etc. With the rising use of nanoparticles in cancer therapy, it is important to collect standardized data and integrate it with machine learning. A field of biomedical informatics called Nano Informatics does this precise function of data standardization and integration. Machine learning algorithms help to predict toxicity, efficacy, potency, etc. In this thesis, an attempt is made to use machine learning techniques such as supervised regression algorithms to predict the efficacy and potency of anticancer nanoparticles. The unsupervised learning algorithms are also used to understand the factors that affect the results. The standard physiochemical and in-vitro data was collected from various scientific papers, listed in the References section. The target variable was cell apoptosis which gives information about efficiency and half-maximal concentration which gives information about potency. The supervised regression models were validated with standard methods like mean square error, root mean square error and R2. The unsupervised learning models were visualized. Both methods were successful in understanding the potency and efficiency and the factors that affect them.