Health matters: How machine learning is helping beat chemotherapy

OncoStem diagnoses chances of breast cancer relapse via data

While machine learning has already proven its worth in many fields, be it travel, logistics or online shopping, health care is emerging as a sector where it is expected to be a game chager. OncoStem Diagnostics, a Bengaluru-based start-up, is using this technology to a greater extent to ease the lives of patients suffering from cancer, especially breast cancer.  

OncoStem was founded in 2011 by Manjiri Bakre, a PhD in Cell Biology from the Indian Institute of Science (IISc), who realised not enough has been done to gauge the aggressiveness of the cancer tumour. 

She felt the necessity of using machine learning to solve this problem out of her personal experience, when she lost a friend who was just 30 to breast cancer, just two years after it was diagnosed.

According to a Lancet report on oncology, breast cancer is the most common type of cancer in India, and continues to increase in incidence. About 150,000 women are diagnosed with breast cancer every year in the country, of which 50 per cent fall under Stage I and II. However, due to lack of recurrence tests, most of them have to go through the painful and toxic process of chemotherapy, even if there is little or no need.

According to Bakre, the procedure (chemotherapy) works in the case of just around 15 per cent of early stage patients. “Chemotherapy is like an atom bomb that kills even the healthy hair cells and blood cells,” she says. OncoStem Diagnostics is solving the problem of over-treatment via chemotherapy through its CanAssist-Breast test. 

The test, developed in over five years by Bakre and her team of 10 innovators, came into being using a cohort of 300 patients. Such procedures are available in western countries, which cost around Rs 300,000 while the typical cost of chemotherapy is around Rs 200,000. 

Hence, most patients opt for toxic therapy rays rather than going for the test that can predict the spreading of cancer to other parts of the body, including lungs, brain, liver and ovaries.

CanAssist-Breast, which is claimed to be the only such indigenous test available in the country, costs around Rs 60,000. It helps 70-80 per cent of the patients who are diagnosed with early stage breast cancer. 

According to Bakre, CanAssist-Breast has been designed especially for Asian patients, as the occurrence of breast cancer in this part of the world happens mostly before menopause, while for those in the Western countries the occurrence takes place after menopause. So, the available western procedures also need to be re-tested in India as patients are much younger here.

The test is offered as a service from OncoSTem’s lab in Bengaluru. It is performed on the patient’s tumour tissue, which the hospital ships to them at room temperature. This proteomic test result is then fed into the company’s machine learning-based statistical algorithm. This, at the end, gives a code that is a number between 1 and 100. If the number is below 15.5, it is a low risk patient and not prone to metastasis.

The Bengaluru firm has so far raised close to $10 million in venture capital funding from Sequoia Capital and Artiman Ventures. Bakre says that almost 70 per cent of the capital raised has gone towards research and development, especially towards building similar test models for other forms of cancer such as oral. These products are expected to be available in the market in 2-3 years. 

The start-up is also planning to come up with triple negative cancer test by next year. The test is currently available for Hormone Receptor Positive breast cancer, and has benefitted about 300 patients in India. 

By 2020, OncoStem is looking to rake in revenue of $4 million. Much of the future growth will be driven by its expansion in to newer geographies such as Singapore, Malaysia and Thailand, which is expected to be concluded by the end of the year. Outside India, the company has launched CanAssit-Breast test in Bangladesh and Sri Lanka.