Bengaluru start-up explores a sustainable drug discovery model for cancer treatment
Premium

immunitoAI is building an AI platform for antibody-based drug discovery which makes designing and developing antibody drugs faster and cheaper

July 21, 2023 08:30 am | Updated 11:16 am IST - Bengaluru

The start-up is currently training its models and has got 80 per cent accuracy in terms of getting biologically viable sequences.

The start-up is currently training its models and has got 80 per cent accuracy in terms of getting biologically viable sequences.

Antibody therapy for cancer treatment, a more precise and lesser harmful alternative to chemotherapy, has been around for many years now. However, for the common person, it has remained prohibitively expensive costing around ₹4 to 5 Crore. Bengaluru-based startup immunitoAI may have a solution for this.

Founded by Dr. Aridni Shah and Trisha Chatterjee, immunitoAI is building an AI platform for antibody-based drug discovery which makes designing and developing antibody drugs faster and more efficient.

Chemotherapy vs antibody therapy

As per reports the market size of antibody therapy is as huge as $186 billion (as of 2021) and is expected to witness a CAGR of 13.2 percent from 2022 to 2028. A relatively new market, Dr. Shah points out that today of the top 10 best sellers in medication, five are antibodies.

Chemotherapy drugs which are small molecule drugs, kill any fast-growing cell, and cannot differentiate between a healthy cell and a cancerous cell. As a result, they harm every fast-growing cell in the body including the hair and the red blood cells.

On the other hand, antibodies bind to a specific target. Making use of this property, pharma companies have been designing antibody-based drugs that can bind to cancer cells. These antibodies would activate the immune system which would attack only the cancer cell.

“Since it is a biological molecule, so far companies have relied on biological sources like animals to get these antibodies,” Ms. Shah explains.

“For example, there are certain proteins specifically present only on cancer surfaces. You can take just that protein and inject it into an animal. For the animal, it’s a foreign body. So, its immune system starts producing antibodies over a course of three months to a year.”

The antibody-making cells are then extracted. Millions of antibodies are screened to identify those that bind to the target. Called lead molecules, drugs are derived from these antibodies.

However, there’s a catch.

The trial-and-error method

“Pharma companies put enormous effort into characterising this molecule. Characterising is about finding out what its properties are, what protein sequence is if it can be stable at different temperatures, and is it possible to store it for a very long time…,” Dr. Shah says. This takes about three to six months.

Next is a trial-and-error stage where mutations are made to the molecule to see if it can attain the required drug properties.

The cycle of characterising, mutating and testing continues until a satisfactory molecule is obtained.

But another challenge here is that there is a limit to mutating the molecule as its source is an animal system. After a point, there are chances that it loses either its function or its structure.

At this juncture, the pharma company has only two options – let go of the project which cost them several years and millions of dollars, or compromise and go ahead.

“If you compromise, there is a higher chance that either in animal studies or human studies the drug would fail,” says Dr. Shah.

AI-designed antibodies

The cause of such limitations is primarily the reliance on a biological source to obtain the molecule, which in the first place was created to be inside an animal body and not to be a drug. But what if antibodies with the required drug properties could be designed with the help of AI and then developed in a lab?

Trisha Chatterjee (left) and Dr. Aridni Shah, co-founders of immunitoAI

Trisha Chatterjee (left) and Dr. Aridni Shah, co-founders of immunitoAI

With this thought process, immunitoAI started building an AI platform with two parts.

The first part called imDESIGN designs a new antibody from scratch with all the required properties. The several thousand trial-and-error procedures over the years and advancements in the field have already generated a huge trove of sequencing data with which the startup has been training its deep learning model.

“We have a generative model which generates a lot of antibodies for a given antigen. Then we take it through a second pipeline called imRANK,” Dr. Shah says.   

“We generate a bunch of antibody sequences, fold it and then dock it to the exact area where we want it to bind. Then we study all the interactions over there based on which they will be ranked as good binders or weak binders.” 

Till here the entire process is done computationally. Now it moves to the lab. DNAs corresponding to selected antibody designs are synthesized and put in either bacteria or mammalian cells. The cells now start producing actual antibodies and characterisation of those antibodies is done.

“Since we have already predicted the drug properties upfront, now characterisation is not an exploratory path or trial and error. At this stage, we simply validate whether the predicted parameters are true or not,” says Dr. Shah.

The startup is currently training its models and has got 80 percent accuracy in terms of getting biologically viable sequences. The next step would be to improve on the binding and specificity.

A sustainable model

According to Dr. Shah once the AI model is trained it would take only minutes to generate antibody designs with the required drug properties. About six to eight months would be required to experimentally validate the molecules in the lab.

She notes that it would take a year or even less to get the final molecule in place. Currently, the industry takes anywhere between four to 10 years to achieve the same results.

While speed is one aspect of it, what is more important is the efficiency.

She says, “Imagine a pharma company that has ten molecules in the pipeline. Each clinical trial causes around $ 500 million. Out of these probably one molecule will finally reach the market. The company would want to make up for all the losses it incurred with the other nine. Then the cost of the drug becomes very high.”

This makes antibody therapy beyond the reach of the common man.

Dr. Shah acknowledges that immunitoAI’s intervention alone may not bring the down prices drastically, but it would certainly increase the success rates significantly, atleast by 2x.

The goal is to ultimately make a more sustainable model of developing the molecules which she hopes would eventually pave the way for making antibody drugs more available and accessible similar to small molecule drugs like crocin.

Challenges

In 2021 the company raised a sum of ₹ 1 million in a seed round led by pi Ventures. Existing investors Entrepreneur First was also part of the round.

Roopan Aulakh, managing director at pi Ventures noted that advancement in technologies like AI is transforming drug discovery and by using AI to accelerate the pace of antibody discovery, immunitoAI has been one of the pioneers in this approach .

“This can translate into a huge impact since antibodies, although very promising candidates for various treatments are constrained by the expensive, long, and inefficient discovery process today. The opportunity that immunitoAI offers is the effective and safe treatment of diseases like cancer and autoimmune diseases through novel antibodies,” Ms. Aulakh said.

“At pi Ventures, we back deep-tech startups leveraging disruptive technologies to change the status quo or create new markets. immunitoAI is well aligned with this vision. They are solving a tough problem and building significant IP in the process,” she added.

The immunitoAI team hopes that the final model would be ready in two to three months and it would then start generating sequences. However, the way forward is not without challenges.

Dr. Shah notes, “In biology, AI has unfortunately been misused. Many have made humongous claims and have not delivered. Hence people are sceptical to accept AI-based solutions in biology.”

She adds that they also often face the question of FDA approvals.

“Eventually a biological molecule will be made and that will go through all the testing FDA requires. But we are coming in the stage prior to that which is about identifying a good antibody. No regulations come there. Everything will come after actual animal testing,” she clarifies.

Road ahead

The team has completed one set of validation and the next target is to prove that the platform can generate antibody designs that bind specifically to the target. 

Dr. Shah says, “We want to benchmark our antibody designs against existing antibodies. We’ll take an existing target that has an existing antibody in the market. And then we will generate ours. This would help to compare how many years it took for the development of the existing drug and how much time it takes for ours.”

She adds, “We are going to look at four to five cancer targets first purely for benchmarking. After that, the pharma companies we partner with would define the disease and the target, and then the drugs would be developed.”  

Top News Today

Sign in to unlock member-only benefits!
  • Access 10 free stories every month
  • Save stories to read later
  • Access to comment on every story
  • Sign-up/manage your newsletter subscriptions with a single click
  • Get notified by email for early access to discounts & offers on our products
Sign in

Comments

Comments have to be in English, and in full sentences. They cannot be abusive or personal. Please abide by our community guidelines for posting your comments.

We have migrated to a new commenting platform. If you are already a registered user of The Hindu and logged in, you may continue to engage with our articles. If you do not have an account please register and login to post comments. Users can access their older comments by logging into their accounts on Vuukle.