The Insights You Need to Develop Precision Medicine for Global Populations
Our Value Proposition:
Better Understand the Populations Drugs Are Being Developed For
Leveraging data-driven analysis, we can understand how different groups respond to medications and utilize those insights to optimize future drug development for those same demographic.
Uncover New Market Segments in Populations Not Targeted with Current Drug Developments
Our algorithms help pinpoint unmet needs in drug development for specific populations, paving the way for targeted therapies in these new market segments.
Identify New Patient Populations to Target with Existing Drugs and Create New Market Opportunities
Post-marketing trials can unlock hidden potential in approved drugs. By verifying off-label uses and identifying new patient populations, we can develop targeted marketing strategies to reach previously underserved markets and deliver these potentially life-saving therapies to more patients.
Our Technology
Grapefruit Biosciences leverages software algorithms that utilizes advanced machine learning models to predict drug efficacy, safety, and response across populations representing diverse genetic ancestries.
Our approach uses expertise in human genetic variation, protein structure analysis, and computational methods to create a suite of powerful tools that will become essential in the drug development process and precision medicine.
GrapefruitBio uses advanced AI to analyze diverse genomes from biobank-scale data and circumvent the laborious clinical trial recruitment process by identifying who should be prioritized for the clinical trial in the first place, or evaluate whether or not a drug should be taken to market.
Reshaping the way you think about drug development by incorporating population genetics.
Why don’t organizations understand the genetics of disease or drug response well?
Limited Genetic Data and Narrow Focus
European Reference Genome: Most current research relies on limited genetic data, focusing on individual genetic variants compared to a reference human genome, which is primarily European.
Genome-wide Association Studies (GWAS): These studies often use genotyping arrays that include only a selection of genetic variants across the genome, ignoring the full spectrum of genetic variation that would be captured by sequencing an individual’s entire genome.
Challenges with Clinical Trial Recruitment
Today, AstraZeneca and other top pharmaceutical giants say they target clinical trial patient diversity samples that match that of the United States, however, two fundamental problems persist:
Laborious Recruitment Process: Reluctance to join clinical trials, stemming from past racial injustices. Additionally, recruitment avenues are not well-established, presenting significant barriers due to a lack of education, access, and awareness.
Knowledge of Basic Health Risks: The current structure of clinical trial recruitment requires individuals to have knowledge of their basic health risks to enroll. For example, last year, 23andMe discovered a genetic variant common among Puerto Ricans that confers a 12-fold higher risk of cataracts. If Puerto Ricans are unaware of their heightened risk, they are unlikely to enroll in cataract clinical trials.
Overly-simplistic Models
Many models assume that a single genetic variant alone influences how an individual’s proteins might interact with a drug, which is an oversimplification.
Pharmaceutical Companies: Many focus on diseases caused by a single genetic variant, making it easier to develop drugs or therapies targeting the aberrant protein produced by this variant. The advent of CRISPR has reinforced this focus by enabling the direct editing of high-impact variants. For example, CRISPR has been successful in treating sickle cell anemia by editing the causative genetic variant (Casgevy).
Penetrance: This term refers to the proportion of individuals carrying a specific gene who express the related trait. It is important to note that other variants and environmental factors can interact with a high-impact variant to result in disease.
Outdated Machine Learning Algorithms
There is insufficient knowledge of disease-causing or drug-response-modifying genetic variants in diverse populations.
Knowledge Gaps: Due to the incomplete biological understanding of disease and drug response in diverse populations, machine learning models trained on standard data sources will not address these gaps. These models will perpetuate biases created by incomplete genomic data, over-reliance on single variants, and the over-representation of Europeans in genetic studies.
Protein Modeling: Efforts to model single protein changes, such as using AlphaFold models, may not adequately capture how natural variations in proteins, which contain several variants in unique combinations, interact with drugs. Modeling naturally occurring protein variations, rather than single changes on a standard reference genome, provides a more comprehensive understanding.
