spsMOSAICS
Why spsMOSAICS
Discover hidden patient subgroups
Identify and map statistically meaningful subgroups that may not be apparent from measured patient characteristics.
​
Identify potential responders
Determine whether identified subgroups differ in their association with treatment, helping reveal potential responder groups.
​
Understand patient heterogeneity
Characterise how patient subgroups differ in risk, frailty, baseline hazard and treatment associations.
​
Quantify subgroup membership
Estimate the probability that individual patients belong to each detected subgroup.
What can spsMOSAICS help you achieve?
Go beyond measured covariates
Traditional approaches to patient stratification often depend on measured characteristics that are already known to be informative.
But what happens when the relevant subgroup structure is not captured by the available covariates?
​
spsMOSAICS uses Bayesian model selection to uncover latent structure directly from clinical data, allowing potentially meaningful subgroups to be identified even when measured covariates provide limited information about subgroup membership
Understand the structure of your patient cohort
Key methodological features spsMOSAICS
-
How many statistically meaningful subgroups are present.
-
How large each subgroup is.
-
How subgroups differ in risk and treatment associations.
-
Which patients are likely to belong to each subgroup.
-
Whether subgroup membership can be predicted from available covariates.
Advanced statistical methodology
Bayesian model selection
spsMOSAICS uses Bayesian model selection to compare the likelihood of possible subgroup structures and determine which provide statistically supported explanations of the observed data.
​
Multiple risks
The pipeline can accommodate multiple risks and account and correct for competing-risk effects when estimating survival-related outcomes.
​
Automated analysis
From subgroup discovery through analysis and reporting, spsMOSAICS provides a fully automated workflow.
Applications
-
Responder identification
Identify patient subgroups with distinct associations with treatment response to support more targeted clinical research. -
Patient stratification
Assess whether a patient cohort can be meaningfully divided into distinct subgroups and characterise the differences between them. -
Biomarker discovery
Explore which measured characteristics are associated with subgroup membership and may help identify potential stratifying biomarkers. -
Risk and survival analysis
Characterise differences in risk, frailty and survival patterns between patient subgroups, including analyses involving competing risks. -
Clinical trial design
Generate insights into cohort heterogeneity that can support prospective responder identification and more targeted clinical trial strategies.
Publications in which the pipeline was used
-
Rowley, M., Garmo, H., Van Hemelrijck, M., Wulaningsih, W., Grundmark, B., Zethelius, B., et al. (2017). A latent class model for competing risks. Statistics in Medicine, 36(13), 2100–2119. https://doi.org/10.1002/sim.7246
-
Barber, P. R., Weitsman, G., Lawler, K. J., Barrett, J. E., Rowley, M. I., Rodriguez-Justo, M., et al. (2019). HER2-HER3 heterodimer quantification by FRET-FLIM and patient subclass analysis of the COIN colorectal trial. Journal of the National Cancer Institute. https://doi.org/10.1093/jnci/djz231
-
Häggström, C., Rowley, M., Liedberg, F., Coolen, A. C. C., & Holmberg, L. (2023). Latent heterogeneity of muscle-invasive bladder cancer in patient characteristics and survival: A population-based nation-wide study in the Bladder Cancer Data Base Sweden (BladderBaSe). Cancer Medicine, 12, 13856–13864. https://doi.org/10.1002/cam4.5981
Are you ready to explore what your data can reveal?
Do you have a challenging biomedical dataset or research question?
​
Whether you are working with a small dataset, high-dimensional measurements, heterogeneous patient populations or distributed clinical data, our team can help explore whether our approach is suitable for your research question.