Personalizing cancer care, one tumor at a time
When cancer knocks at one’s door, pretty much like a dreaded guest, one doesn’t know the extent of the dread immediately. As the guest settles in, the unpleasantness unravels. It’s somewhat similar feeling with cancer diagnosis. One doesn’t know if it’s a deadly form of cancer and in what stage; what will be the prognosis; and if one is lucky enough to convert the cancer into a manageable chronic disease or if it’s going to be a death sentence.
For very long, scientists have been attempting to nail the diagnosis part of it using biomarkers and genomic sequences so that at least doctors don’t experiment with a cocktail of toxic drugs but give more targeted therapy. Neither approach has proven to be foolproof or supremely effective. The presence of a biomarker does not necessarily translate into a beneficial clinical outcome.
Now, a significant advancement in accurate prediction of drug regimen is reported by a young biotech company in Bangalore. Nearly six years after being set up in Boston and then shifted to Bangalore, Mitra Biotech has developed a new technology platform, CANScript, which handles much of the complexity in predicting which cancer drug will be effective in which patient. To use a pharmaceutical jargon, it’s a First-in-Class diagnostic that renders truly personalized cancer diagnosis – one tumor, one patient at a time.
Mitra Biotech reported its findings in Nature Communications on February 27. Its chief scientific officer, Pradip K Majumder, says what his team has been able to address through this technology is to identify the driver of the cancer motorcade. It literally is a motorcade: cancer is a combination of many diseases where thousands of oncogenes are involved, most of them mere passengers at the mercy of the driver. If you identify the driver, you can halt or alter the speed of the motorcade.
For instance, the popular and blockbuster drug trastuzumab, known by innovator’s brand name Herceptin, is used in breast cancers where a particular biomarker, Her-2 gene, is over expressed. So what we know today is that if this biomarker is over-expressed (or Her2 positive), then trastuzumab is the silver bullet. What is less known, especially among non-scientists, is that not all Her-2 positive cancers respond well to this drug. Majumder says published data shows that as many as 50 percent of patients receiving trastuzumab may not be effectively benefiting from this drug. The reason, using his driver-passenger analogy, is that even though Her-2 is over expressed, it may not be the driver in that particular patient.
The same applies to another biomarker KRAS (which is a predictor for colorectal cancer). If this gene is mutated then the common drug cetuximab will not work. But even if the gene is not mutated, or is present as the wild type, then too it’s not certain if that the drug will work. “We collected global data and found that cetuximab is effective only in 10-20 percent cases of colorectal cancer,” says Majumder.
As this 2013 study shows, biomarkers as a key tool for tumor identification and treatment determination have many limitations and pitfalls.
Biomarkers were the subject of hot pursuit in the late 90s and early 2000s. After that, when DNA sequencing became mainstream following the dramatic reduction in sequencing cost, exome and whole genome sequencing formed the next wave of personalized cancer treatment. Micro array and sequencing data taught us much about the biology of gene expression, but good clinical outcomes remained elusive though a handful of pioneering companies like Foundation Medicine in Boston did provide some succor to patients. (Earlier this year, Roche, the largest maker of cancer medicines, paid $1.03 billion to Foundation for 56.3 % stake.) The company wasn’t doing too well and one of the reasons was that running cancer gene panels, and not strongly impacting the clinical outcome in terms of which drug to use, wasn’t exciting enough for insurers as well as hospitals.
Mitra’s CANScript has gone a step further. A simpler analogy would be the bacteria sensitivity tests that are commonly used today. Just as a pathology lab takes a swab, cultures it and tests it against all available antibiotics to finally help a doctor prescribe the right antibiotic, CANScript runs a test against the biopsy from the patient and gives a score card for the drugs to be used. In clinics it is currently used in six solid tumors (breast cancer, gastrointestinal, glioblastoma, head and neck squamous cell carcinoma and colorectal) and two blood cancers. Three other cancers – lung, cervical and melanoma—are under lab testing.
However, the limitation with CANScript is that it requires very fresh tumor. That is, the biopsy (or the surgically removed tumor) must reach Mitra’s lab within 20 hours. Multiple labs will be required for the testing to reach more care givers. The company today makes the organ transplant box and the buffer available to the medical oncologists who then ship the biopsy sample along with 5 to 10 ml of the patient blood. The sample on reaching the lab undergoes various assays and stays in the incubator for 3-4 days during which the tumor is challenged with various drugs. The results, or functional readouts as they are called, are collected and fed in the computer model which then gives a score card for 4 to 6 drug combinations.
In short, the technology mimics the human tumor ex-vivo. The machine learning algorithm allows the platform to compile all the different parameters of tumor behavior (cell death, cell proliferation, cell viability, etc) into a linear score reading – less than 25 makes the drug ineffective, more than 60 is a good bet.
Frankly, the true profitable market for the product resides outside India but the company has begun reaching out to 2000 odd medical oncologists in the country. One of the early adopters is HCG hospitals which has been part of the early clinical testing of this product.
Nearly one in eight persons in India will have cancer by 2020, according to the World Health Organization. Unarguably, it’s precision diagnostics that can cut cost and morbidity for patients.
PS: Since some readers asked about the paper, here’s the title and the URL:
Predicting clinical response to anticancer drugs using an ex vivo platform that captures tumour heterogeneity