After a decade of collaborative development with colleagues at Boston’s and Cambridge’s leading academic institutions, Biomodels has enabled an innovative, unique analytical method for genomic microarray data that produces meaningful, translatable, and actionable outcomes and information. Our proprietary computational algorithms are based on the development of Bayesian Networks to describe gene‐relationships that impact phenotypes and have five primary applications:

1. Defining mechanism(s) of action for drugs and biologicals associated with efficacy endpoints associated with a range of clinical indications including cancer supportive care, inflammatory diseases, metabolic diseases, and drug toxicities-both clinically and in animal models.
2. Differentiating and defining responder/non-responder populations in clinical trials (particularly phase 2).
3. Validating the predictive value of animal models as being representative of human drug/biological response.
4. Identifying at risk populations for treatment-related toxicities.
5. Rescue/resuscitation of failed phase 2 clinical studies.
The BIOMODELS’ approach differs from classic candidate gene or genome‐wide association studies in five important ways:

Old Approach Biomodels’ Bayesian Network Approach
Hypothesis‐driven:Investigator pre‐selects genes from known pathways (which accounts for only a small portion of the array output) No pre‐selection/hypothesis/prejudice: entire gene/SNP array output is analyzed and learned networks lead to conclusions
Genes evaluated one at a time with a preset threshold of significance All genes are evaluated for their interactions with one another‐No preset thresholds
Final model based on logistic regression and the best fit to outcome of interest Final model based on cross‐validated gene clusters to arrive at cluster which has the greatest predictive power
Often difficult to replicate‐many false positives Data is robust and easy to replicate
Known genes/pathways‐difficult to generate novel ideas and/or intellectual property Results are typically not intuitive or obvious‐Easy/fast to generate new targets and/or intellectual property

BIOMODELS’ genomic analysis approach is a reality and has been successfully applied to array outputs generated in animal and human studies.


• SNP-Based Bayesian networks can predict oral mucositis risk in autologous stem cell transplant recipients.
• Personalized medicine for mucositis: Bayesian networks identify unique gene clusters which predict the response to gamma-D-glutamyl-L-tryptophan for the attenuation of chemoradiation-induced oral mucositis.
• Gene expression changes in peripheral blood cells provide insight into the biological mechanisms associated with regimen-related toxicities in patients being treated for head and neck cancers.