Researchers at Monash University's Biomedicine Discovery Institute have developed an AI model that translates one type of molecular data into another while preserving spatial context, potentially lowering the cost and technical barriers of multi-omics research. The model, named NicheTrans, was described in a paper published on 15 July 2026 in Nature Methods.
Bridging the gap in spatial biology
Scientists increasingly use multi-omics approaches to understand disease by combining information from genes, proteins, and other molecular signals. Acquiring multiple high-resolution data types from the same sample is often expensive and technically challenging. NicheTrans addresses this by learning relationships between different molecular data types and using those relationships to accurately infer missing information. It is the first computational approach specifically designed for spatially-aware cross-omics translation.
"Because many important omics measurements are often expensive and technically difficult, NicheTrans provides a game-changing solution," said Professor Jiangning Song, head of the AI-driven Bioinformatics and Biomedicine Lab at Monash BDI and co-senior author of the study. "It accurately fills in missing data, making multi-omics analysis highly accessible and affordable."
How NicheTrans works
The model uses transformer-based deep learning to infer biological information, such as protein expression, from more readily available data like gene expression measurements. Crucially, it incorporates spatial context, allowing researchers to study not only which cells are present but how they are organized and interact within tissue environments.
During validation across multiple biological systems, NicheTrans identified spatial patterns and relationships that single data types alone could not detect. The software has been released as a free, open-source resource for academic and non-commercial research.
Applications in cancer and brain disease
Understanding cellular neighborhoods and tissue architecture is becoming important for complex diseases such as solid tumour cancers, Alzheimer's disease, and Parkinson's disease. "NicheTrans can reveal important architectural relationships between different cells, or the association between different biomarkers," Professor Song said.
In cancer research, the tool may help identify specific cell populations linked to treatment resistance or relapse. In Alzheimer's, it can characterize how different cell types are organized across brain regions. The researchers believe the approach could support more precise patient stratification by linking spatial biological data with clinical outcomes.
Professor Song is keen to apply NicheTrans to analyse solid tumour samples from patients who relapse after treatment, hoping to identify biomarkers and cellular signatures associated with tumour recurrence.
Why this matters for science and research professionals
NicheTrans lowers the barrier to advanced spatial biology by enabling labs worldwide to generate high-quality multi-omics insights from more affordable technologies. For scientists looking to build AI skills, resources like the AI Learning Path for Research Scientists provide practical training in machine learning and data analysis. The open-source release also creates opportunities for collaboration with pharmaceutical and biotechnology companies, potentially accelerating clinical translation.
Your membership also unlocks:
