See all projects

Predicting Melanoma Patient's Responses to Nivolumab Immunotherapy Using Machine Learning Models

Hello! My name is Dnyanada Vijapure, and my Polygence project is on early cancer detection with AI. I chose to work on this project because I would like to help people who are suffering from this deadly disease. After my project is complete, I would like to continue studying this topic and make a career in it.
High School
Monta Vista High School
Graduation Year
Student review

View Polygence scholar page
Project description

Melanoma is a cancer that takes over melanocytes, cells that give skin pigment and causes them to multiply continuously. This can lead to health problems because if melanoma starts spreading it can go into organs in your body and stop them from functioning properly. One immunotherapy that helps boost the immune system against melanoma skin cancer is an IV-administered drug called Nivolumab. Nivolumab is used to block the connection between PD1 and PDL1. PD1 is a protein on T cells and if connected to PDL1, a protein on cancerous cells, it allows the T cells to stop killing other cells including cancerous cells. The number of genes that are expressed tells whether the patient is getting affected negatively or positively. Although not all patients respond well to Nivolumab, we can hypothesize that gene expressions tell whether a patient responds well or not to treatment. In a recent experiment obtained from a website called, it was found that the melanoma patients who responded well to Nivolumab had specific patterns of gene expression. The patient's reactions included whether the disease was progressive, a partial response, a complete response, or a stable disease. Next, using the scikit-learn library and supervised learning algorithms such as nested logistic regression models, a machine-learning model was created to determine if the patient is likely to be benefited from using Nivolumab. The average accuracy of this model was 65%. The purpose of this research is to help doctors find out whether a patient with melanoma skin cancer will have a positive outcome if they are given Nivolumab. It would allow doctors to confidently notify their patient’s prognosis. For the patients that are not predicted to have a good response, doctors can decide to give an alternate therapy.

Predicting Melanoma Patient's Responses to Nivolumab Immunotherapy Using Machine Learning Models
Project outcome

The outcome of this project was the identification of several genes which were significantly expressed using machine learning models such as logistic regression. It included following genes: ATP6V1E1P2, CST3, ED3, F2L, PCMT1, TAF1, WT2, NELFE, TBL3, Fras1, Cdc37, Fgl2, Ifnar2, Igk-V10A, mea, Ngf, Rsp4, and Shh. These results provide valuable insights for healthcare professionals in determining a patient's response to Nivolumab immunotherapy.

MD/PhD Doctor of Medicine and of Philosophy candidate
Comp Sci, Medicine, Biology, Engineering
Machine learning for clinical applications, Protein engineering, Immunoengineering, Cancer immunotherapy, Medicine, Biology, Computer Science
Mentor review

They helped me with fixing and answering any questions or problems I had.

Interested in starting
your own project?
Apply today!

Already have an account? Log In
By registering you agree to our terms of use and privacy policy, and consent that we or our partner provider may reach out to you using a system that can auto-dial.