Email
Password
Remember meForgot password?
    Log in with Twitter

article imageQ&A: AI fighting cancer by rescuing and repurposing failed drugs Special

By Tim Sandle     Sep 19, 2019 in Science
Lantern Pharma is aiming to change how the industry thinks about clinical testing, by analyzing historical data with AI and machine learning to get the right people in the right trials with the right drugs.
The pharmaceutical sector is spending billions on cancer research, but most cancer drugs fail clinical trials. Such medications are not one-size-fits-all solutions, but they are often tested as if they are, with people getting treatment they may be genetically indisposed to responding to. Lantern Pharma has developed an AI-driven platform to predict which types of tumors are sensitive to specific drugs, so trials with the right patients can yield more successful results.
Digital Journal spoke with Panna Sharma, CEO of Lantern Pharma to discover more about the technology.
Digital Journal: Why do most cancer drugs fail clinical trials?
Panna Sharma: Oncology drugs are notorious for having high rates of failure, especially in late stage clinical trials, with less than four percent of drugs developed for oncology making it to market. In recent years, oncology clinical trials’ success rate has notably improved due to the use of heightened patient stratification, biomarkers to better sub-type the cancer, and improved clinical trial design to accommodate the changing landscape of precision-driven cancer treatment. No single organization or group is to blame, but the fault lies with clinical trial design lacking methodologies and approaches to segment patients based on the molecular or genomic characteristics of the tumor.
As new classes of drugs come to market, they also continue to raise the standard of care and improve outcomes, making it more challenging to gain approvals. Each form of cancer has different types of tumors or different subtypes of that cancer and a high degree of variability in how said cancer may respond to one drug or a combination of drugs. For this reason and many others, each of those tumors can react differently to the same drug, rendering the drug ineffective…to some, not effective enough…to others, and effective to a small subset of patients. By understanding the biological mechanisms and underlying genomics of the responsive patients, we can improve drug selection, trial design and patient outcome.
By designing trials to match the right therapies to the patients that have the highest potential response rate, those very same drugs deemed ineffective in the general population may be proven to be significantly effective in a targeted subpopulation, resulting in better patient outcomes (saved lives). Until recently the genomic data, technologies and methodologies to solve these complex prediction models, underlying drug prediction, were not available in a scalable and cost-effective manner. Technologies that are making machine-learning, cloud computing and large-scale genomics accessible and available are significantly and positively impacting the failure rates of cancer drug development.
DJ: How important is precision medicine for treating cancer?
Sharma: Precision medicine will completely change the way cancer is treated, and that is already beginning to manifest itself in the scientific and clinical community. For decades, oncologists have worked hard to specialize in individual cancers, emphasizing the unique characteristics and complexities of each tumor type based on the location and pathology of the tumor. However, precision medicine brings to the forefront the idea that each tumor can be unique, and tumors in very different locations can respond to the same drug based on the similarity of the genomic profile of the tumor, and not the pathology of the tumor.
This fundamental premise of precision medicine is that the molecular and genetic basis of the tumor will dictate how the tumor behaves, what drugs it may respond to, how aggressive it might be, and potentially if it will generate multiple clones. Based on the patient’s clinical history, we can also predict and prepare for the patient becoming resistant to a cancer drug and preemptively change or combine therapies to improve the patient’s outcome. Additionally, new technologies, such as CAR-T and Immuno-Oncology, are enabling us to use the patient’s own immune system or T-cell repertoire to help fight off or mitigate the cancer. These approaches along with the wider adoption of genomic and transcriptomic profiling of patients will improve outcomes and reduce cancer care costs.
DJ: How important is assessing historical data in the quest for new cancer drugs?
Sharma: In addition to classifying drugs as ineffective, the high failure rate of oncology drug trials is costly, as each failed trial expends financial, medical and scientific resources. Historical data can provide oncologists and scientists with a better sense of a drug’s efficacy before clinical trials, making each trial more efficient and reducing the time and cost of bringing drugs to market. Bringing precision drugs to market quicker and more cost-effectively benefits patients by improving their access to oncology drugs in more ways than one. The continued understanding of genomics and how they relate and cause patients to respond to cancerous tumors will only accelerate, ultimately improving outcomes for cancer patients for years to come.
Past clinical trials obtained biological samples from patients to draw conclusions about drug efficacy and determine a biological profile of the patients and tumors. At the time, science and technology had not advanced enough to draw conclusions about genetics, and certainly whole genome and RNA sequencing were not available. It is possible to go back through and gather genomic data from clinical trial samples and correlate them with drug efficacy.
DJ: How can artificial intelligence assist with this?
Sharma: Artificial intelligence can analyze quantities of highly complex data that humans aren’t capable of analyzing. In oncology, artificial intelligence can be used to analyze the results of millions of past clinical trials to determine which types of tumors are most likely to positively respond to certain drugs. Knowing that, oncologists can design their clinical trials with the optimal combination of cancer tumors and drugs so that the drugs can be as effective as possible with those tumor populations. Further, AI enables clinical trials to proceed much quicker and provide decisive outcomes, a welcome sight for patients, practitioners and the investment community.
In past clinical trials, genomic historical data may have been gathered, and the AI and machine learning tools were not advanced enough to draw conclusions about genomic markers. This technology is now readily available, and this past data can be retrospectively analyzed.
DJ: What is the Lantern approach?
Sharma: Lantern’s proprietary AI platform, the Response Algorithm for Drug Positioning and Rescue (RADR), is intended to personalize the use of cancer drugs by identifying what genetic features are most likely to make the tumor respond to our drug (and potentially other cancer drugs). We are currently using this AI platform to help uncover and create robust genomic signatures that can be used to accelerate the development of our portfolio as well as reduce the cost of our clinical trials by more accurately predicting the exact type of patient that will have the highest likelihood to respond to our portfolio of drugs.
Lantern has developed our portfolio of oncology therapies using AI. By rescuing “failed” drugs, RADR helps create outcomes through:
Assembling and analyzing big data (genomic, clinical, response),
Identifying patient subgroups through machine learning and AI,
Clarifying mechanisms of action,
Identifying potential combinations,
Helping reestablish and relaunch a faster and more efficient path to the clinical trial setting.
RADR offers important benefits in retooling new oncology drugs, including identifying new patient populations for failed or abandoned drugs based on genomic stratification and selection, shortened time to market, reduced risk in development, potential for orphan or fast track status based on developing a precision pathway.
DJ: What successes have you had?
Sharma: RADR has helped fast track a once-shelved asset (a DNA damage repair inhibitor that showed promise in a number of clinical trials) from in-license to out-license within 18 months. The core of this success was the ability to highlight and identify the genomic profile of patients with metastatic, hormone refractory prostate cancer that were most likely to respond to this drug, LP-100. Additionally, RADR is being used to identify the key genetic driver across a variety of solid tumors that correlate to the potential for high response to another one of our drugs, which also works by inhibiting DNA-damage repair in tumors.
This identification of the key genetic driver and a potential gene signature would have taken 12 to 24 months with a fairly large team in traditional drug development, but with our data-driven approach augmented with biological insight and curating relevant literature, took only three months.
DJ: How are clinical trials progressing?
Sharma: Our most advanced program, LP-100, is actively recruiting patients with metastatic prostate cancer in Denmark and Germany. We are also using RADR, to conduct a clinical study of the drug LP-184, which will, then, help guide the design of a precision clinical trial that will be guided by a genomic signature.
Finally, we are actively involved with analyzing the genomics behind LP-300, which was a drug that failed a large scale, global phase 3 trial. We are actively reviving LP-300 for a targeted trial in non-small-cell lung cancer for women that have never-smoked. For context, failed female non-smokers had survival rates improve by 125% - to nearly two and a half years - over standard of care through a combination of chemotherapies. We expect to review the design of the trial, improve the target patient population and resurrect the trial using a genomic-guided, precision approach.
More about Artificial intelligence, Cancer, Cancer treatment, Medication
More news from
Latest News
Top News