ORAONCO integrates next gen sequencing with machine learning to detect oral cancer at a cellular level. Could you walk us through how the vector matrix algorithm was developed and how it learns from the Indian phenotype metadata?
Over the last five years, the scientific team at Genefitletics has been working tirelessly to elucidate the deep rooted connection between systemic inflammation & mitochondria dysfunction, the root cause of all forms of cancer including oral & throat cancer. Departing from the existing dogma of cancer being a genetic disease caused by random genetic mutations, we focussed on the molecules secreted by oral microbiome and their interactions with mitochondria as a focal point to breakdown the cellular mechanism that damages mitochondria and triggers oral and throat cancer.
The data related to oral microbiome and their biochemical transactions with mitochondria was so huge that there was a need to build an algorithm that works at the intersection of biology, chemistry and mathematics to decipher and make sense of the data in order to understand pathogenic process in the onset of oral cancer and deconvolute those processes into early molecular targets.
Based on the existing scientific literature, we built an initial cancer biology model that helped us build predictive biomarkers for oral & throat cancer. We then carried out internal validation studies on five billion molecular data points from our existing Indian molecular database related to downstream functions of the oral microbiome to test the hypothesis that enabled us to identify early molecular signatures of oral cancer. Our algorithm underwent 100 plus iterations before it arrived at a final model that combines & quantifies 27 biochemical pathways using proprietary single value decomposition & vector matrix methods (machine learning models) to detect oral cancer at an earliest stage.
Molecular data coming from every new sample by sequencing the saliva sample is churned & mined by the model to further fine tune the algorithm to improve the efficacy.
The model also mines the phenotype metadata coming from every new subject to analyse the impact of the interventions which acts as a feedback loop to further improve the effectiveness of interventions.
You’ve mentioned five billion molecular data points underpinning the test. From a systems biology perspective, how do you ensure the relevance and clinical accuracy of such a vast dataset when applied to real-world Indian patients?
Five billion molecular data points from our existing Indian population specific repository of molecular data was classified into model development (healthy controls) & validation based on specific symptomatic conditions used to identify the subjects carrying oral and throat cancer.
The oral cancer detection mathematical algorithm was developed using a 2X2 vector matrix model on data of 3 billion molecular data related to healthy controls. The model so developed was applied to all the subjects classified into model development cohort (healthy controls) to detect the specificity. Based on the analysis, the specificity on healthy control was 95 per cent which established a clinical accuracy with statistical significance of P value (Adj) < five per cent.
As a final step, the model so developed was applied on 2 billion molecular data points related to the validation cohort (those carrying symptoms related to oral cancer) to detect the presence of oral cancer. This final analysis was able to establish a sensitivity of 70 per cent with P value (adj) <0.05.
This clinical accuracy and efficacy is going to increase further & finetune the algorithm as we feed the model with more molecular & phenotype data with every new customer sign up.
In a real world setting, ORAONCO, our oral cancer early detect test, utilises next generation sequencing to sequence 10 million molecules from an individual’s saliva sample and overlay the proprietary algorithm, built on five billion molecular data sets, to translate 10 million molecular data into 27 biochemical pathways that detect oral cancer before the physical examination or biopsy show up any signs
How does ORAONCO differentiate itself from conventional oral cancer screening methods in terms of sensitivity, specificity, and predictive value?
The current clinical practice for diagnosing oral cancer depends upon genetic mutations, physical examination and expertise of healthcare practitioners which generally look at downstream epiphenomenon, post cancer specific organ (mouth) symptoms, observable lesions in mouth & visible tumor size in PET scans. These evaluation methods fail to identify cancer in early stages and cannot go deep into change in biochemistry that triggers cancer onset, growth & tumor proliferation.
Systemic inflammation & oxidative stress causing mitochondria dysfunction that leads to oral cancer development are asymptomatic and initiates years sometimes decades before the post cancer symptoms show up. These changes are generally missed by current investigations, leading to late diagnosis. Most of the oral cancer cases (>70 per cent) are detected in stage III or IV when the tumor has progressed/grown or even lead to a situation of metastasis. At these stages, the five year survival rate is less than 50 per cent. ORONOCO is leading the early detection from the front by identifying and measuring these Inflammatory changes decades ago by looking at cellular pathways causing mitochondrial dysfunction, leading to 4X better predictability, 70 per cent sensitivity and 95 per cent specificity. This translates to an> 85 per cent five year survival rate.
India contributes to nearly a third of global oral cancer incidence, yet clinical pathways remain outdated. How do you envision ORAONCO shifting the clinical protocol for oncologists and dentists in India?
ORAONCO is a quantum leap forward in redefining how cancer is viewed and analysed. The current clinical practice relies heavily on irrelevant DNA & healthcare delivery data which results in less than 3 per cent therapeutics success.
With focus on molecular data & feedback loop using longitudinal metadata, ORAONCO envisions to give broader access of relevant clinical insights related to oral microbiome and mitochondrial dysfunction and related success stories to oncologists and dentists.
We plan to regularly publish clinical studies, case studies and success stories in relevant journals to create a data driven level playing field for the healthcare practitioners to shift & embrace to new emerging clinical protocols that could make a paradigm shift in the treatment of oral cancer.
Given the scale of the oral cancer burden and current diagnostic gaps, what’s your roadmap for ORAONCO’s integration into hospital systems or diagnostics labs at scale? Are you looking at partnerships with government or private oncology chains?
ORAONCO currently works on a direct to customer model to offer its early detection solution covering detection and cancer modifiable nutritional therapeutics interventions. We plan to collaborate with cancer specialty hospitals, oncologists and private oncology chains to offer ORAONCO both for preventing the onset and recurrence of oral cancer.
For preventing recurrence, ORAONCO will blend its nutritional therapeutic interventions with standard of care therapies offered by hospitals to first starve the cancer cells (having intratumoral & intratumoral heterogeneity) of the fermentable fuels and eliminating the balance cancer cells (having genetic homogeneity) using chemotherapy.
With increasing interest in precision medicine, how do you see the role of biotech startups like Genefitletics shaping India’s preventive healthcare landscape in the next decade?
Over the last five decades, the current healthcare system has been focussed on collecting and driving healthcare decisions based on healthcare delivery data that gives a post mortem of biological processes taking place inside the body. However, interventions based on this healthcare delivery data have delivered limited therapeutics success and at best have managed the symptoms of diseases. The healthcare practitioners lack access to relevant datasets to understand the pathogenesis of disease at an early stage.
With wide scale adoption of exponential technologies, the focus is shifting from healthcare delivery to molecular data at least in developed economies. This novel practice is being implemented in India.
Biotech startups such as Genefitletics could lead the revolution in preventative healthcare from the front by collecting, analysing and interpreting these molecular data sets to fill in the precision gap and provide the healthcare practitioners with access to these never measured molecular data, empowering them to detect the diseases at an early stage, measure the effectiveness of their existing treatments & make appropriate modifications in their treatment protocols to prevent, stop the progression of disease or outrightly reverse them.
neha.aathavale@expressindia.com
nehaaathavale75@gmail.com