Since the earliest days of medicine, physicians have needed to make judgements about a patient’s condition. In his Book of Prognostics, Hippocrates (after whom the Hippocratic Oath is named) wrote of a doctor: “He will manage the cure best who has foreseen what is to happen from the present state of matters.”

Foresight is never easy, but thankfully disease diagnosis and prognosis has improved since Hippocrates’ time. As doctors deepened their understanding of diseases, they identified measurable signals, or biomarkers, that indicate normal biological function, pathology or response to treatment. For example, body temperature is a biomarker for fever while your blood glucose level is a marker for diabetes. Today, biomarkers can be measured down to a molecule. These types of signals have increased in their importance to medical diagnosis and prognosis.

In the past decade, a new set of technologies have emerged that dramatically accelerated the discovery new biomarkers. The work promises to reshape how doctors diagnose and treat disease and how pharmaceutical companies pursue drug development. 

In oncology, the impacts are already being felt. In the past 18 months, the FDA approved the first and second tissue-agnostic therapies. The drugs target solid tumors that show specific biomarkers—namely microsatellite instability (MSI) and neurotrophic receptor tyrosine kinase (NTRK) gene fusion—irrespective of where the tumors are growing in the body. For a field that has been anchored to anatomy in the diagnosis of disease, a biomarker-first approach represents a sea change. 

Beyond cancer, researchers are investigating a pro-peptide known as PRO-C3 as a potential biomarker for nonalcoholic steatohepatitis (NASH), which currently requires an invasive liver biopsy for diagnosis, as well as anti-citrullinated protein antibodies (ACPA) that could help diagnosis of rheumatoid arthritis before joint damage occurs. 

In a panel discussion held in New York City, leading researchers gathered to explore how biomarkers research—coupled with cutting-edge technology like artificial intelligence and digital pathology—could improve patient care and change the way diseases are diagnosed and treated. Scientific American’s Custom Media Group, in partnership with Bristol-Myers Squibb, hosted the event. According to one attendee, Suzanne Topalian, associate director of Johns Hopkins Bloomberg-Kimmel Institute for Cancer Immunotherapy, “If you had to think of one topic that should be our focus today in advancing oncology, it is biomarkers.”


While biomarkers have been fundamental to medicine for ages, the development of molecular biomarkers as therapeutic targets is relatively new. One of the first and best examples was the approval of Herceptin in 1998. The drug targets overexpressed estrogen receptors that can be predicted by the Her-2/Neu oncogene, which was discovered more than a decade earlier.

The success of Herceptin, along with the advance of molecular biology techniques and DNA sequencing, led to an explosion of biomarker research. Today, a few dozen molecular biomarkers are considered clinically relevant, and a handful, namely PDL1, ALK, EGFR, CTLA4 and NTRK, have become foundational therapeutic targets.

Those successes have been the exceptions, not the rules. Cancer is incredibly complex. The scientific community has only recently amassed a significantly greater understanding of how tumor biology and the tumor microenvironment play a role in the development of disease and response to therapy. In addition, reliance on traditional study designs, complex statistical methods and a lack of reproducibility slowed the movement of basic biomarker research from the lab to the clinic.

The situation today is different. With the introduction of new technologies, “The pace of innovation in biomarker discovery is going to rapidly take off,” said Saurabh Saha, M.D., Ph.D., senior vice president and global head of translational medicine for Bristol-Myers Squibb. Translational medicine is an interdisciplinary research approach that “translates” basic biomedical research into clinically relevant tools and medicines. Saha’s group is working on new ways to profile a patient’s specific disease biology in order to provide patients with tailored therapeutic strategies.

Several technologies are accelerating biomarker discovery. Next generation genetic sequencing, along with high-throughput transcriptomics and proteomics, now permits scientists to amass huge amounts of data that was previously inaccessible. The growth of single-cell biology could help scientists expand those data sets even further, giving scientists an unprecedented understanding of certain diseases.

With so much data on hand, the development of new big data techniques also serves as an accelerant. Computational biologists and bioinformaticians use them to transform mountains of data into relevant and significant signals. The adoption of machine learning will only speed the movement from raw data to valuable insights. 

“At its core, biomarkers are a machine learning problem,” said Andrew Beck, CEO of PathAI, a Boston-based startup that aims to advance pathology with machine learning and deep learning techniques. “We're actually trying to predict the optimal treatment for a given patient based on all the data you can extract from that patient.”

Left to right: Moderator Fred Guterl and panelists, Saurabh Saha, Suzanne L. Topalian, Lynette M. Sholl,and Andrew H. Beck discussing biomarkers research. Credit: Scientific American Custom Media


For all the forces that are driving biomarkers research forward, others hold it back. Data is often siloed within organizations or inaccessible to the larger scientific community. It is also often collected with few standards in mind, so scientists cannot blend data sets. At the event, researchers remarked that the cost of validating a biomarker for clinical use is still exceptionally high, meaning that work is mostly underwritten by large pharmaceutical companies – or not at all. This has led to a situation where speculative discovery work has far outpaced clinically relevant validation.

Yet while acknowledging those challenges, panelists agreed that the forces favoring the development of clinically relevant biomarkers are now more significant than those against it. That shift carries with it big implications.

“We can use the molecular data to go back and reclassify a lot of the diseases that we’ve put into boxes over the years,” said Lynette Sholl, associate professor of pathology at Harvard Medical School. “We’re realizing that we can actually look down the scope and potentially even predict what the driver mutation could be based on the way a tumor looks.”

Beyond applying existing biomarkers to known diseases, researchers will be advancing more biomarkers out of discovery. “We’re leveraging information for biomarkers such as tumor metabolism, the microbiome, the presence of viruses, and oncogenic viruses that drive certain cancers,” Topalian said.

And these biomarkers can be combined for greater precision. “There’s going to need to be a quantitative blend where each [bio]marker has an incremental contribution,” Beck said. “The aggregate understanding of all those markers is ultimately going to give the physician more of an understanding of what exactly is going on in that patient’s tumor.” 

Multiplexing will invariably require machine learning, since the data sets will be massive and computation heavy. And that adoption of machine learning will carry implications of its own. Beck said that in the future, pathologists will be empowered by artificial intelligence (AI) to be even better at their job, taking care of some of the low-level problems that computers are good at and allowing pathologists more time and resources to focus on some of the more complicated cases. On the same note, AI could relieve doctors from certain decisions and enable them to focus more on patient care, which is, ultimately, why this is all happening in the first place.

“The expectation is that it will allow us to be much more precise about counseling patients and selecting the right patients for the right treatments,” Topalian said.

While that expectation once seemed distant, it no longer does. According to Saha, from Bristol-Myers Squibb, “If you take all the technology we have today and marry that with the medicines that are currently available and in development, the future looks incredibly bright. I’m more confident than ever in our ability to leverage translational medicine, across the full spectrum of the drug discovery and development process, to drive innovative scientific advancements for some of the most challenging diseases facing patients now and in the future.”

To learn more about biomarkers research at Bristol-Myers Squibb please visit