Google Cloud launches AI tools to accelerate drug discovery

Google Cloud on Tuesday launched two new AI-powered tools aimed at helping biotech and pharmaceutical companies accelerate drug discovery and advance precision medicine.

One tool, called the Target and Lead Identification Suite, is designed to help companies predict and understand the structure of proteins, a fundamental part of drug development. Another, the Multiomics Suite, will help researchers ingest, store, analyze and share reams of genomic data.

The new development marks Google’s latest advancement in red-hot AI armor racewhere tech companies are vying to dominate a market that analysts believe could one day become worth trillions. The company has faced pressure to showcase its generative artificial intelligence technology since the public launch of OpenAI’s ChatGPT late last year.

Google announced its generative chatbot Bard in February. Shares in parent company Alphabet rose 4.3% last week after Google made the announcement several advancements in AI at its annual developer conference.

The two new Google Cloud suites help address a long-standing problem in the biopharma industry: the lengthy and costly process of bringing a new drug to the US market.

Pharmaceutical companies can invest anywhere from a few hundred million dollars to more than 2 billion dollars to launch a single drug, according to a recent report from Deloitte. Their efforts are not always successful. Drugs that reach clinical trials have a 16% chance of being approved in the US, says another Deloitte report.

The high cost and dismal success rate is accompanied by an extensive and tedious research process that usually lasts approx. 10 to 15 years.

The new suites will save companies a “statistically significant” amount of time and money throughout the drug development process, said Shweta Maniar, Google Cloud’s global head of life sciences strategy and solutions. Google did not provide CNBC with specific numbers.

“We help organizations get medications to the right people faster,” Maniar told CNBC in an interview. “I’m personally very excited, this is something that me and the team have been working on for a few years now.”

Both suites are generally available to customers starting Tuesday. Google said the cost will vary by company. Several companies, including Big Pharma’s Pfizer and the biotechnology companies Cerevel Therapeutics and Colossal Life Scienceshave already used the products.

The target and lead identification suite aims to streamline the first key step in drug development, which is identifying a biological target that researchers can focus on and design a treatment around, according to Maniar.

A biological target is most common a protein, an important building block for disease and all other areas of life. Finding that target means identifying the structure of a protein, which determines its function, or what role it plays in a disease.

“If you can understand the role, the protein structure and the role, now you can start developing drugs around that,” Maniar said.

But that process is time-consuming and often unsuccessful.

Researchers can take around 12 months only to identify a biological target, according to a generally followed guidance for drug manufacturers published in a database operated by the federal National Library of Medicine. The two of them technician scientists traditionally use to determine protein structures also have a high rate of failure, according to Maniar.

She also said it’s difficult for traditional tech to increase or decrease the amount of work they do based on demand.

Google Cloud’s suite takes a three-pronged approach to making that process more efficient.

The suite allows researchers to ingest, share and manage molecular data about a protein using Google Clouds Analytics Huba platform that allows users to securely exchange data between organizations.

Scientists can then use this data to predict the structure of a protein with AlphaFold2a machine learning model developed by a subsidiary of Google.

AlphaFold2 runs on Google’s Vertex AI pipeline, a platform that enables researchers to build and deploy machine learning models faster.

In minutes, AlphaFold2 can predict the 3D structure of a protein with more accuracy than traditional technologies and at the scale researchers need. Predicting that structure is crucial because it can help researchers understand a protein’s function in a disease.

The final component of Google Cloud’s suite helps researchers identify how the protein’s structure interacts with different molecules. A molecule can become the basis for a new drug if it changes the function of the protein and ultimately shows the ability to treat the disease.

Researchers can use Google Clouds high performance computing resources to find “the most promising” molecules that could lead to the development of a new drug, according to a press release about the new tools. These services provide infrastructure companies need to accelerate, automate and scale up their work.

Cerevel, which focuses on developing treatments for neuroscience diseases, typically has to screen a large library of 3 million different molecules to find one that will produce a positive effect against a disease, according to Chief Scientific Officer John Renger. He called that process “complicated and involved and expensive.”

But Renger said the company will be able to sift out molecules more quickly using Google Cloud’s suite. Computers will handle the screening of molecules and help Cerevel “get an answer really quickly,” he said.

Renger estimates that Cerevel will save at least three years on average by using the program to discover a new drug. He said it’s hard to estimate how much money the company will save, but emphasized that the suite cuts down on the resources and manual labor typically required to screen molecules.

“What that means is we can get there faster, get there cheaper, and we can get drugs to patients much faster without as many failures,” he told CNBC.

Cerevel has been working with Google for more than a month to further understand the suite and determine how the company will use it. But Renger hopes Cerevel will “be in a place where we get some results” in the next month.

Google Cloud’s other solution, the Multiomics Suite, aims to help researchers tackle another daunting challenge: genomic data analysis.

Colossal Biosciences, a biotechnology company that aims to use DNA and genetic engineering to reverse extinction, has used the Multiomics Suite in its research.

As a startup, Colossal did not have the internal infrastructure required to organize or decipher massive amounts of genomic data. A human genome sequence alone requires more than 200 gigabytes of storage, and researchers believe they will need 40 exabytes to store the world’s genomic data by 2025, according to National Human Genome Research Institute.

The institute estimates that five exabytes could store every word ever spoken by humans, so building the technology to support genomic data analysis is no small task.

As such, the Multiomics Suite aims to provide companies like Colossal with the infrastructure they need to make sense of large amounts of data so they can spend more time focusing on new scientific discoveries.

“If we had to do everything from scratch, I mean, that’s the power of Google Cloud, right?” Colossal’s vice president of strategy and computational science, Alexander Titus, told CNBC in an interview. “We don’t have to build it from scratch, so it definitely saves us time and money.”

Scientists’ ability to sequence DNA has historically outstripped their ability to decipher and analyze it. But as technology has improved in recent years, genomic data has unlocked new insights into areas such as the genetic variations associated with disease.

Google Cloud’s Maniar said that could ultimately help in the development of more personalized medicines and treatments. In 2021 alone, two-thirds of drugs approved by the Food and Drug Administration were supported by human genetics research, according to an article published in the journal “Nature.”

Maniar believes the Multiomics Suite will help encourage further innovation.

Ben Lamm, CEO of Colossal, said the Multiomics Suite is the reason the company has been able to perform research on “any reasonable timeline.” Colossal began testing Google’s technology late last year, and as a result, Lamm said the company is on target to produce a woolly mammoth by 2028.

Without the Multiomics Suite, Lamm said he believes the company would have been set back by over a decade.

“We wouldn’t be anywhere near where we are today,” he said.

Before using Google Cloud’s suite, much of Colossal’s data management was done manually using spreadsheets, Lamm said.

He said it would have been a “massive burden” for the company to try to build the more complex tools needed for research.

“We are no longer in small data when it comes to biology,” Colossal’s Titus said. “We’re thinking about the scale of how do we get insights into 10,000, 20,000, 10 million years of evolutionary history? And those questions simply aren’t answered without scalable computing infrastructure and tools like cloud computing and Multiomics.”

Correction: It can take about 12 months for researchers to identify just one biological target, according to a widely followed guidance for drugmakers published in a database operated by the federal National Library of Medicine. An earlier version had the typo.