The first question most people have when talking about artificial intelligence is if it will help us or harm us. But in the health industry, the answer to that question is clear. Whether it is being used to identify patients at risk or interrogate X-rays and scans to aid in diagnosis, AI promises to save lives and reduce costs.

CytoReason is one company using AI to help in the war against various ailments and conditions, and it has discovered a unique biomarker that can predict who will and will not respond to anti-TNF drugs.

What are anti-TNF drugs, otherwise known as TNF inhibitors? They’re a type of drug used to treat inflammatory conditions such as rheumatoid arthritis, psoriatic arthritis, juvenile arthritis, inflammatory bowel disease (including Crohn’s and ulcerative colitis), ankylosing spondylitis, and psoriasis.

Founded in October 2016, and located in Tel Aviv, Israel, CytoReason has been working with Haifa’s Rambam Healthcare Campus and Tel Aviv Sourasky Medical Centre to apply its AI to help improve efficiency.

Using an AI model of the immune system, CytoReason identified patterns in the genomic data of patients who responded to these drugs versus those who did not. Comparing the results of the two groups, the company has discovered specific biomarkers in blood and tissue samples that can predict who will respond to anti-TNF drugs such as infliximab (sold as Remicade).

“What the AI allows us to do is to identify certain features (biomarkers) that are common among patients who respond to the Anti-TNF,” CytoReason cofounder and president David Harel told me. “Once a biomarker is found, the determination whether the patient might respond can be done almost instantly. The true benefit of the AI is its ability to absorb big data and within minutes to answer questions, such as — in this case — what are the common features to patients who respond to a specific treatment?”

The results appear impressive. While outlined in more detail in a recent study, the biomarkers were validated with an 82 percent tissue biomarker accuracy (for patients with an inflammation score over 2.5) and 94 percent blood biomarker accuracy.

AI isn’t just offering the opportunity to shorten the process for Remicade treatments — CytoReason is using it to accelerate identification of the efficacy of other therapies too. “In this particular case, AI was used to speed up the search for biomarkers,” Harel said. “In many other cases, it is used to speed up the discovery of new drugs.”

Does AI make it entirely possible to determine the efficacy of an anti-TNF treatment alone, or are patient trials still needed?

“We are discovering biology, not inventing it,” Harel said. “This is the major distinction that exists between the application of AI in life sciences and most other fields. That suggests that every new technique introduced can and should be validated in the laboratory or through trials, but these trials will not look the same as the ones we are used to. AI is changing the purpose and method of clinical trials.”

In other words, AI is helping change the landscape of how clinical trials are run.

“Since AI allows us to make precise predictions, the validation of these predictions requires short, less-expensive trials,” Harel said. “Today, and in particular in cancer research, we can see trials that use only a fraction of the traditional number of patients, and this trend is only in its infancy.”

Of course, AI can’t speed up the clearance of a particular type of drug or treatment, and it doesn’t offer a chance to leapfrog any regulations that are in place from drug administrations and other regulatory bodies.

“It is important to remember that the entire domain is strictly regulated, and only in the past few years the regulators have grown comfortable with the idea,” Harel said. “The approval of treatments, based on computational markers, is a new practice, less than five years old.”

The biomarkers identified in today’s announcement are just the beginning, and the results of the AI’s ability to determine which patients will respond to drugs like Remicade are backed by additional analysis and study.

“We ‘asked’ our model how the immune system differed between patients who responded to the treatment and those who did not,” Harel said. “Upon identifying the differences, we ran two 52-patient trials in two separate hospitals to validate our results. Finally, we ‘asked’ again what are the ‘levers’ that control these variations in the immune system and can be traced in the blood. We validated that result in another, third 22-patient trial. Overall, we recruited 74 patients to the validation trials, in three independent groups.”

The speed difference AI is offering here is quite notable. “Typically, the process to determine whether a patient was responsive to anti-TNF drugs like Remicade took over three months and was dependent on trial and error, exposing the patient to potential side effects and unnecessary costs,” Harel said. AI, in this case, can provide an answer in minutes.

Not only does AI give immunologists a system-level view of immune responses, it also cuts costs, time, and risks for pharmaceutical companies during their development cycle. In addition to helping prevent safety issues, CytoReason’s AI could create a more personalized medical experience and reduce unnecessary prescriptions.

So what’s next for CytoReason’s technology, and will it expand to other types of treatment?

“We are already using our AI platform to discover biomarkers in the most popular autoimmune treatments and cancer immunotherapies,” Harel said. “As our model ‘sees’ more data, it will be able to predict growing number of treatments in a growing number of therapeutic areas. It is being tested now on certain dermatology, cardiovascular, and neurological treatments.”