Helping to Untangle Cancer Drug Resistance with Data

Share this post:

Why do targeted cancer therapies often fail? We have acquired so much more understanding about cancer in the last fifty years than in the last five thousand years. Approaches to patient treatments have dramatically changed, and statistics show significant improvement in patient response and outcomes to therapy in the last half a century [1]. Yet one of the biggest remaining barriers to cancer treatment is our limited understanding of how cancer cells acquire resistance to the drugs used treat them.

To help better understand this complex issue and potentially lay the groundwork for more effective treatment, IBM and the Broad Institute launched a research initiative in 2016 aimed at discovering the basis of cancer drug resistance. Drawing on data and computational methods, our intent is to help researchers understand how cancers become resistant to therapies, which is currently not very well understood.

We are now publishing one of the first significant results of this collaboration in the scientific journal Nature Medicine. This specific research looks at the tremendous amount of information held in our DNA and genetic data, and what it could tell us about cancer, how it evolves and how it could be treated. Specifically, we used computational data mining methods to examine clues held in DNA data – acquired from blood samples – to shed new light on acquired drug resistance in gastrointestinal cancers.

Gastrointestinal cancers comprise some of the most serious and complex diseases today, including pancreatic, liver, colon and stomach cancer. Across our combined team of researchers, computational biologists and clinicians, we uncovered three critical elements surrounding drug resistance:

  • Blood biopsies can provide more complete information about a tumor’s potential drug response than data gleaned from much more invasive tissue biopsies
  • Patients can acquire multiple resistance mechanisms – this will be the first time that this finding has been shown in such a large-scale study. This helps to explain why targeted therapies can often fail.
  • Data algorithms can help untangle the most difficult cases in understanding why a patient has acquired drug resistance, potentially laying the groundwork for the use of this technology to build more personalized treatment plans.

What is tumor evolution?

Ironically, acquired resistance to drugs emerges in the same way that the disease itself emerges – via the evolution of cell populations. It is intriguing that the forces that drive the evolution of a population of cells (tissue or tumor) are the same as those that drive the evolution of a population of single-cell bacteria or multi-cellular higher order organisms (plants, animals), at least in a mathematical sense.  This gives us the advantage of being able to use similar quantitative models to study this phenomenon. The underlying principles and algorithms used to uncover clues about the human population can be reused and reengineered to answer questions about tumor evolution.

When cancer progresses in a patient and metastasizes to different locations in the body, whether in the same organ or a different tissue, lesions are formed in these sites. Important information about the patient’s cancer can be obtained from biopsying the detectable lesions and used to make informed treatment decisions. However, tumors or lesions are heterogenous and can exhibit different molecular profiles when obtained from different tissues, or even from within the same tissue [2, 3].

What is acquired resistance?

Simply speaking, when a patient is observed to be responding to a drug (for example, shrinking tumor size or in remission), and the disease abruptly stops responding, it is envisaged that the patient’s disease has acquired resistance to the drug. One current theory of resistance is that a minority of tumor cells with the ability to withstand the assault of the drug exist at low levels in the heterogenous population of cancer cells and begin to dominate the population following the treatment that is effective against the majority [4].

These cells are characterized by mutations which can be identified as “mechanisms,” in clinical cancer parlance, of the developed resistance. It is still unclear whether the resistance mechanism is intrinsic (i.e., it already existed in a few cells prior to therapy) or acquired (i.e., it developed de novo as a result of the treatment).

Our recent research makes the following three discoveries:

1. Blood captures more tumor heterogeneity than biopsies of tissue lesions

Capturing the heterogeneity contained within and across all tumor lesions in a single patient, though desirable, is very difficult. Many lesions are too small to detect or unreachable for biopsy. Our study demonstrated that blood biopsies, which examine cell-free DNA (cfDNA) in the blood plasma, contain more genomic information describing the heterogeneity of a patient’s lesions than standard clinically obtained tissue biopsies. Thus, the simpler procedure of drawing a blood sample, as opposed to a more invasive incision, can provide more information about the disease.

2. Multiple acquired resistance mechanisms in a patient

This is the very first study of this size showing the acquisition of multiple resistance mechanisms in individual patients. The standard of care today in many cases operates under the assumption that resistance in a patient will eventually be driven by a predominant mechanism, and new treatment regimens that aim to address resistance are geared towards one or two specific targets [5]. The presence of multiple resistance mechanisms in a single patient could potentially explain the inefficacy of preventing resistance. This also suggests that the use of multiple targeted therapies simultaneously, each associated with a different set of resistance mechanisms, may result in a more desirable response.

3. Data Mining Algorithms help uncover acquired resistance mechanisms in difficult cases

In some patients, it is difficult to determine the precise acquired resistance mechanisms from the molecular profiles provided by either blood or tissue biopsies. Using domain-specific, detailed models of tumor evolution, along with pattern discovery, we were able to suggest potential resistance mechanisms in some of these difficult patient scenarios. Thus, clever algorithms extract more clinically relevant information than can be directly observed in the patient data.

As one of our initial publications, we’re excited for the foundation that this study lays for future research into acquired drug resistance, both in gastrointestinal cancers and across other cancers as well. We look forward to seeing how this research takes root and expands, as well as to obtain additional results that we expect to publish in the coming months.

Pattern Discovery Algorithms enable the elicitation of Mechanism of Resistance (MoR) for the difficult cases based on the patients with known MoR.

  4. Burrell, R.A. & Swanton, C. Tumour heterogeneity and the evolution of polyclonal drug resistance. Oncol. 8, 1095-1111 (2014).

IBM Fellow, IBM Research

Gad Getz

Institute Member, Broad Institute

Ryan Corcoran

MD, Assistant Professor of Medicine at Harvard Medical School

More Healthcare stories

Advancing the Potential of AI in Medical Imaging at MICCAI 2020

I believe one of the most promising areas for AI to make an impact is in the field of medical imaging. Through advancements in AI that allow for more intelligent and accurate analysis of video and still images, there is hope that clinicians will soon be able to widely augment the data and information they […]

Continue reading

AI Data Tracker Encourages Scientific Research into COVID-19 Non-Pharmaceutical Interventions

What impact do measures such as shelter-in-place, mask wearing, and social distancing have on the number of COVID-19 cases? How do the COVID-19 quarantine measures that have been implemented by North American countries compare to South American countries? These are just a few questions about the wide range of non-pharmaceutical interventions (NPIs) that have been applied by governments, globally.

Continue reading

IBM Research and the Broad Institute Seek to Unravel the True Risks of Genetic Diseases

In 2019, IBM and the Broad Institute of MIT and Harvard started a multi-year collaborative research program to develop powerful predictive models that can potentially enable clinicians to identify patients at serious risk for cardiovascular disease (1, 2). At the start of our collaboration, we proposed an approach to develop AI-based models that combine and […]

Continue reading