How SMART Research is Shaping the Future of Health, Technology, and AI

At SMART, our interdisciplinary research groups (IRGs) are focused on driving innovations to tackle critical challenges of societal concern. Over the past few months, SMART has delivered breakthroughs on several fronts, including tools to understand cell biology, antimicrobial resistance, disease management, and the future of AI and work.

Advancing human health through innovative technology

RNA modifications play a critical role in regulating how cells function, respond to stress and adapt to disease, but understanding these molecular changes has remained challenging due to limitations in existing analytical tools.

To address this, researchers from SMART's Antimicrobial Resistance (AMR) IRG have developed a first-of-its-kind RNA-specific isobaric tandem mass tagging platform that enables the identification, localisation and quantification of RNA modifications with high precision. Adapted from techniques commonly used in proteomics, the platform integrates sample preparation, mass spectrometry analysis and data processing into a single workflow.

Using the platform, the team identified and quantified 24 RNA modifications in Pseudomonas aeruginosa transfer RNA (tRNA), including two modification sites not previously characterised in bacteria. The findings provide new insights into how RNA modifications change across different genetic and environmental conditions, creating new opportunities to advance RNA biology research and accelerate future diagnostic and therapeutic developments.

The need for faster, more accessible tools also applies to human health applications. The trillions of microbes living in the human gut influence everything from inflammation to metabolism, yet the tools available to monitor gut health biomarkers are often costly, time-consuming and confined to specialised laboratory settings, making them impractical for routine screening or point-of-care use.

Researchers from SMART's Disruptive & Sustainable Technologies for Agricultural Precision (DiSTAP) IRG, alongside its collaborators, have developed a fluorescent nanosensor capable of rapidly detecting indole-3-propionic acid (IPA), an emerging biomarker linked to gut health and disease.

As the first optical nanosensor specifically designed to detect IPA, the platform provides a rapid optical readout within minutes and can distinguish IPA from closely related metabolites commonly found in the gut. Validated using 125 human plasma samples, the sensor successfully revealed significant differences in IPA levels between healthy individuals and patients with inflammatory bowel diseases, including Crohn's disease and ulcerative colitis.

Better health outcomes also depend on the materials that make long-term monitoring and care possible. Hydrogels are widely used in biomedical applications ranging from wound dressings to wearable health monitors because of their softness and high water content, but conventional hydrogels struggle with limited air exchange efficiency, posing bottlenecks in long-term use.

Researchers from MIT and SMART Wearable Imaging for Transforming Elderly Care (WITEC) IRG have developed a new class of air-permeable hydrogels using a technique known as viscoelastic phase separation. By creating a stable network of air-rich aerogel particles within a water-rich hydrogel, the team achieved record-high oxygen permeability while maintaining approximately 70% water content. The resulting material combines breathability, flexibility and biocompatibility, helping prevent fluid accumulation, improving skin comfort during prolonged wear and supporting stable physiological monitoring.

Mapping the health of communities

Urban design and environments of cities play a critical role in shaping population health outcomes. Researchers from SMART's Mens, Manus and Machina (M3S) IRG, in collaboration with the National University of Singapore (NUS), asked why some neighbourhoods consistently produce better health outcomes than others, and analysed health, geospatial and urban data across more than 28,000 census tracts in the United States.

Published in Nature Health, the study found that equitable access to urban services, green spaces and well-connected neighbourhoods is strongly associated with improved physical, mental and overall health. The research also revealed that improving urban services in lower-income communities could deliver disproportionately large health gains. By identifying the urban characteristics most closely linked to health outcomes, the study provides new evidence that can help policymakers and urban planners build healthier, more equitable cities.

Wastewater surveillance has also emerged as a powerful tool for tracking infectious diseases across entire communities. However, most existing approaches focus on detecting genetic material from pathogens, providing limited insight into how much immunity a population has built through previous infections or vaccinations. In a study published in Water Research in May this year, researchers from SMART's AMR IRG explored whether antibodies found in wastewater could help fill this gap. 

The team found that antibodies are abundant in wastewater but are often fragmented into smaller components that can still recognise pathogens, and that wastewater proteases pose a major obstacle to detecting them accurately.

By developing improved methods that account for antibody fragmentation and protease activity, the researchers demonstrated that population-level immunity signals can be detected and measured from wastewater samples, opening up new ways of monitoring population-level immunity.

Shaping the future of AI and work

As AI adoption accelerates, governments, businesses and workers increasingly face the question of which tasks can be meaningfully supported or transformed by AI technologies. To better understand AI’s real-world reach, researchers from M3S explored this challenge by mapping more than 13,000 AI software applications and 20.8 million robotic systems against a newly developed ontology of approximately 40,000 work activities. 

The study revealed that AI software is concentrated in "Think" activities such as writing, coding and analysis, while robotics is primarily deployed in "Do" activities such as manufacturing, cleaning and transporting goods. The findings suggest that AI software and robotics are largely complementary technologies rather than direct substitutes, offering new insights into how emerging technologies will reshape the future of work.

The M3S team has also tackled one of the biggest challenges and a core limitation of large language models (LLMs): their knowledge becomes outdated almost immediately after training, and retraining is costly and resource-intensive. Their proposed framework, called MeMo: Memory as a Model, encodes new knowledge into a dedicated memory model that works alongside an existing LLM, so AI systems can use new information without modifying the underlying model itself. 

This allows updates without costly retraining and avoids issues such as catastrophic forgetting. The approach has the potential to support complex, multi-document reasoning while remaining resilient to noisy retrieval environments.