Google: AI is more profound than fire and a key to the future of life science R&D

Google CEO Sundar Pichai has often highlighted the transformative impact of AI, stating that it’s a “more profound” tool than the human discoveries of fire or electricity. Shweta Maniar, Global Director of Healthcare and Life Sciences at Google, echoed this sentiment in a recent interview. AI’s applications in the life sciences are broad, and the technology is gaining ground in everything from drug discovery efforts to workflow optimization. While adoption is early, Google sees big potential.

The tech can help improve healthcare outcomes, helping improve “billions of lives,” Maniar said. Additionally, she sees AI tools as a catalyst to address long-standing challenges in the life sciences, helping potentially chip away at the high costs and lengthy timelines associated with traditional drug development processes. Its customers include Bayer, Gingko Bioworks and the startup Superluminal Medicine.

A recent Bloomberg report also pointed out that Google is tapping AI, in this case trained on 300 million audio clips “including coughs, sniffles, sneezes and breathing” to detect diseases like tuberculosis through sound analysis, further underscoring AI’s expanding role in healthcare. 

Generative AI has the potential to significantly cut time and cost, “bringing new therapies to market faster to the right patients at the right time,” Maniar said. Its uses are multifaceted. The technology can:

• Accelerate the identification of promising drug targets, shortening the early stages of drug discovery.
• Optimize clinical trial design, recruitment, and analysis, leading to faster and more efficient trials.
• Generate synthetic data to augment limited real-world datasets, enabling research in areas with scarce data, such as rare diseases.
• Design novel molecules with desired properties, opening up new possibilities for drug development.
• Predict drug interactions, helping researchers design safer and more effective drug combinations.

Synthetic data, tangible gains in life sciences

While data, along with breakthroughs in compute and algorithms, is a driver of the current wave of AI interest, clean real-world data can sometimes be scarce. Synthetic data can help overcome the paucity of annotated medical data in areas such as rare diseases or specialized oncology indications.

For instance, a 2023 study published in Clinical Cancer Informatics found that AI-generated synthetic data can effectively mimic real clinical and genomic features and outcomes in hematologic malignancies while also anonymizing patient information. The study, which GenoMed4All and Synthema European consortia led, showed that synthetic data can help resolve hurdles related to lack of information, data augmentation, and class imbalance, ultimately accelerating research and precision medicine in hematology.

Maniar highlighted the potential of generative AI to create synthetic data that mirrors real-world datasets. This approach, she explained, can significantly accelerate drug development. “We’re exploring how to augment or support organizations developing medicines in areas with limited real-world data,” Maniar stated.

“Bayer is now working with us on synthetic images for oncology, created from histological images,” Maniar said. “There’s limited data and information available to train algorithms, particularly in the rare disease space.”

This synthetic data approach not only helps in rare diseases and oncology but also has broader applications in clinical trials. When applied strategically, the use of synthetic data not only enhances the robustness of AI models but also leads to tangible gains.

The multifacted potential of multi-modality

The application of AI in life sciences is expanding beyond traditional data formats to multi-modality. “We can now look at processing documents that aren’t just PDFs. They can include charts, audio, and video,” Maniar said. This multimodal approach enables life sciences companies to process and analyze a wider range of data types across their operations, from clinical trial documentation to patient records and research data.

In drug discovery, generative AI is opening up new possibilities. Maniar notes that organizations are exploring “molecule generation models to design novel molecules with desired properties, then expanding into the chemical space for drug candidates.” This advanced modeling capability extends beyond molecule design. Maniar also highlights that “biotech can use models to predict how different drugs interact with each other within the human body,” enabling scientists to explore “designing safer and more effective drug combinations.”

Overcoming obstacles in drug development

Despite advancements in technology and research, the pharmaceutical industry continues to face significant challenges in bringing new therapies to market. Maniar acknowledged these difficulties, stating that “bringing new medicines to market is not easy. We’re seeing continued challenges.”

One major hurdle is the increasing cost and complexity of drug development. “It’s a costly endeavor, with increased regulatory hurdles and supply chain challenges that have come to light in recent years. R&D costs are rising,” Maniar explained. “We need to do more with less, faster.”

The drug discovery process itself is fraught with complexities and potential delays. “Even finding a suitable biological target involved in a disease that can be tackled by drugs can take up to a year, according to the NIH,” Maniar noted.

Clinical trials, a crucial step in drug development, are also prone to challenges. “We’re seeing human error in document generation for clinical trials and challenges around protocol development,” Maniar pointed out. “It’s costing organizations more money, an

“It’s costing organizations more money, and they’re losing out on time and resources. The same goes for trial design, where some challenges or minor mistakes can be extremely costly.”

Furthermore, the industry faces mounting pressure to accelerate drug development timelines. Maniar highlighted the “challenges of increased cost of these lengthy developments and frankly, a world that is no longer going to accept the proverbial 20 years to get a medicine to market.”

Finally, limited data availability, particularly in areas like rare diseases, poses a significant obstacle for AI development and training. Maniar emphasized the “limited data and information available to train algorithms, particularly in the rare disease space.”

Reshaping drug discovery and delivering hope

The application of AI in life sciences is already demonstrating tangible impact, particularly in accelerating drug discovery and streamlining critical processes.

Maniar highlighted the immediate benefits of generative AI, noting significant progress in “speeding up the search for promising drug targets” and improving various aspects of clinical trials, including design, deployment, recruitment, and analysis.

Beyond these immediate applications, AI is also showing promise in enhancing other stages of drug development, such as predicting drug effectiveness, optimizing clinical trial execution, and generating synthetic data to address data limitations in specific research areas, especially those with limited real-world data like rare diseases.

Maniar emphasized the long-term transformative potential of AI, stating that “generative AI has the potential to significantly reduce time and cost, bringing new therapies to market faster to the right patients at the right time.”

This potential is particularly evident in AI’s ability to tackle complex challenges, such as understanding brand new molecules with specific therapeutic properties. Maniar believes that “generative AI models” in this area have “the potential to unlock new drug classes and treatment options we haven’t even considered.”

“This is not just about incremental improvements,” Maniar concluded. “AI has the power to fundamentally reshape the drug discovery landscape and accelerate the development of life-saving therapies for patients in need.”