Now Reading
How This Biotech Company Integrates Machine Learning Algorithms To Detect Relapse Of Breast Cancer

How This Biotech Company Integrates Machine Learning Algorithms To Detect Relapse Of Breast Cancer

Srishti Deoras

Dr Manjiri Bakre who was then pursuing a PhD at Indian Institute of Science had a friend who was diagnosed with breast cancer. The cancer was detected early on and she underwent surgery soon after. While Dr Bakre and her friend rejoiced at the successful treatment, cancer unfortunately recurred in her body at 2-3 sites, while she was pursuing her post-doctoral fellowship. The cancer was aggressive and despite all the treatment options, she succumbed to it within 2-3 years of diagnosis. 

This unfortunate turn of events led Dr Bakre to think that there needs to be more awareness about the course of cancer. She strongly felt that knowing about the aggressiveness of cancer whether it is a ‘small or big’ tumour can help patients plan their treatment and life accordingly. Even if there were tests in the Western countries, they were out of reach of patients because of the cost or different biology in Indian patients. 

This led her to found OncoStem Diagnostics in 2011 for Indian patients to take tests that can predict the recurrence of cancer, just like the Western counterparts. 



In an interaction with Analytics India Magazine, Dr Bakre shared that OncoStem’s first focus has been the development of CanAssist-Breast which identifies the risk of recurrence in patients with Hormone Receptor-Positive Breast Cancer. Similar tests in Oral, Lung, and Colorectal Cancer are under development. 

What Is CanAssist

CanAssist Breast falls under the class of prognostic tests. It is performed only on Hormone receptor-positive (HR+) and HER2 negative breast cancer patients in the early stages. The test is performed on the patient’s surgically removed tumour sample which is preserved in the hospital laboratories. When a patient is diagnosed with early-stage breast cancer, there is a dilemma in the doctor’s mind whether the patient will benefit from chemotherapy or not. Through this test, they can overcome this dilemma. 


W3Schools

CanAssist Breast essentially looks at the biology of the tumour and how aggressive it is. It has been shown in multiple studies that two patients with the same sized tumours, one patient may not have cancer recurrence whereas the other patient may have recurrence. The difference in these two patients is not the tumour size but the tumour biology. That’s the assessment CanAssist Breast provides.

CanAssist uses machine learning to predict this. On asking how she merged cell biology into machine learning, she said that it was her interest in keeping updated with the latest technologies that led to this merger. “When I was evaluating statisticians to work on CanAssist Breast, I was keen to explore machine learning because of its suitability to cancer research. Machine learning is very well suited to study complex datasets and identify patterns,” she shared.

How Does It Use Machine learning-based Algorithms 

Dr Bakre explains that the standard treatment for HR+ HER2-negative breast cancer involves both chemotherapy and hormone therapy. The risk of relapse in early stage (Stage I and II) HR+ HER2-negative breast cancer is very low (10-15%) even if patients are given hormone therapy alone. This implies a majority of patients (~85%) are being overtreated with chemotherapy, which has toxic side effects and lowers the quality of life of patients.  

There is a need to identify exactly which patients benefit from chemotherapy, helping the remaining ~80% patients avoid this toxic treatment and its side-effects.  

OncoStem has worked with 10 hospitals in India to make CanAssist Breast a reality. Tumor tissues were analysed for multiple biomarkers/proteins reflective of aggressive biology of the tumor.  Using this data support vector machine (SVM) based statistical model was developed which assigns a ‘risk score’ based on the key selected five biomarkers and clinicopathological information for each patient. The ‘risk score’ is indicative of risk of recurrence. It categorizes patients based on the ‘risk of cancer recurrence’ clearly as either ‘low or high’.  

“Thus, by analysing a patient’s tumor CanAssist Breast ‘identifies’ patients who will have minimal benefit (Low-risk) or who will benefit the most (high risk) by adding chemotherapy to their treatment,” she said.

CanAssist Breast is validated in India, US and Europe in a multi-centric validation study. OncoStem’s “Made in India” test is most suitable for Indian Patients. Various international quality related and regulatory approvals (NABL, CE-IVD, ISO13485) were obtained, before CanAssist Breast was commercially launched in 2016.

Machine Learning Tools At OncoStem

OncoStem’s machine learning experts work with various machine learning techniques like Support Vector Machines (SVM) with linear and Radial Basis Function (RBF) kernel, Random Forest (RF),  Elastic Net (ESL), multilayer perceptron (MLP), and normal mixture modeling. 

“Our experience with using machine learning in breast cancer has shown us that it is more accurate compared to traditional methods being used, and are strictly focused on maximizing diagnostic accuracy. The “transfer functions”, used in ML are also generally more flexible, allowing them to model complex processes like cancer prognosis,” she shared.

OncoStem’s Growth Story

CanAssist Breast has been in the market as a prescribed test for the last one year. “Today, we have about 200 oncologists who are actively prescribing this test and we are currently talking to more than 1200 clinicians regularly to create awareness about the test,” said Dr Bakre. They have also added more validation data, increasing the confidence of doctors in the test and increasing the number of prescriptions. 

“We have launched patient assistance programs to ensure that patients don’t lack access to the test for financial reasons. Our test is also now reimbursable by health insurance companies in India, removing one more barrier to use. Our laboratory has various international accreditations and we take pride in finding ourselves in the company of a limited set of labs in India who have been able to mark tremendous growth,” says Dr Bakre. 

The Way Forward

Having raised a total of~ $9M from two investors, Sequoia Capital and Artiman Ventures, the funds were primarily used over the last six years to develop and validate the test in the laboratory and to get all the global accreditations to launch the test in India and near India countries. 

The team is currently working on a second test for breast cancer and tests for oral and colorectal cancer. Research is underway towards identifying and characterizing novel drug targets for breast and oral cancer.

“We are working on full automation of our test CanAssist Breast by developing digital pathology solutions that will allow complete automation and also decentralization. This will increase throughput and also allow any hospital in the world to conduct the testing in their own laboratory,” shares Dr Bakre before signing off. 

Provide your comments below

comments

Copyright Analytics India Magazine Pvt Ltd

Scroll To Top