Cancer is as individual as fingerprints—so why do treatments often follow a one-size-fits-all approach? To achieve better patient outcomes, researchers are embracing digital tools to predict how individuals will respond to cancer therapies.
“Much like a vast puzzle, living cells contain pieces of interconnected information from the expression of thousands of genes, proteins and metabolites,” highlighted Kumar Selvarajoo, a Senior Principal Investigator at the A*STAR Bioinformatics Institute (A*STAR BII). These complex connections hold vital clues to treatment outcomes.
However, current computational methods tend to focus narrowly on a single data type, such as specific mutations in tumours, overlooking broader factors including a patient’s overall health and medical history. This fragmented approach makes it difficult to predict who will benefit from a particular treatment, Selvarajoo explained.
To tackle this challenge, Selvarajoo and Md Mamunur Rashid, a Scientist at A*STAR BII, developed MOMLIN—a machine learning framework that integrates diverse biological data to form a comprehensive cellular profile.
Selvarajoo compared MOMLIN to a master gardener: it doesn’t focus on just one type of information, such as the colour of petals (clinical data). Instead, it takes a holistic approach, factoring in the shape of the leaves (genetics), the quality of the soil (cellular environment) and even the weather conditions (other biological influences).
MOMLIN was built by integrating data from 147 breast cancer patients—including genetic mutations, RNA expression and clinical attributes—and used advanced algorithms to identify key multimodal biomarkers and predict treatment outcomes with improved accuracy.
When validated, MOMLIN outperformed existing tools, achieving 98.9 percent accuracy in drug-response classifications. Using MOMLIN, the research team also discovered that treatment success was associated with robust immune responses, such as increased CD8+ T-cell activity, while resistance was linked to immune evasion mechanisms like T-cell exclusion. These findings highlight new therapeutic targets to counter drug resistance.
Selvarajoo said that MOMLIN provides a richer understanding of cancer biology, and by uncovering the molecular mechanisms behind treatment resistance, could pave the way for more effective, personalised therapies.
The team has already filed a technology disclosure license for MOMLIN and is working to expand its applications to other cancers and diseases. Future plans include testing drug combinations and using single-cell atlases to investigate how individual tumour cells respond to treatments.
“Ultimately, we want to make MOMLIN more accessible to researchers and clinicians worldwide, by developing a user-friendly and scalable platform that can be applied to any disease,” Selvarajoo concluded.
The A*STAR-affiliated researchers contributing to this research are from the A*STAR Bioinformatics Institute (A*STAR BII).
