Hugo Saavedra
CEO & Co-founder
Chih-Ming Ho, one of KYAN’s scientific co-founders, pioneered work in aerospace engineering. In the 1990s, he developed microelectromechanical systems (MEMS) that could actively sense and control turbulence around aircraft — a problem once thought intractable [1] . By embedding tiny sensors and actuators directly into the airframe, his team showed that small, real-time interventions could stabilize enormously complex, chaotic systems. The lesson was clear: to control large forces, think small.
This same systems-engineering principle underpins Optim.AI™. Cancer biology, like turbulence, is nonlinear, unpredictable, and resistant to brute-force approaches. Instead of trying to model every molecular detail or rely on massive data sets, Optim.AI™ applies Small Data artificial intelligence (AI): we measure how a patient’s living tumor cells functionally respond to various drugs in real time, and from these compact but highly informative datasets, we extract the intervention points that matter most. Small, precise drug perturbations — identified through functional sensing — can redirect the behavior of an entire tumor ecosystem.
In short, Optim.AI™ extends the same philosophy from turbulence to tumors: with the right, highly sensitive sensor technology and small, targeted interventions, even the most complex systems can be steered toward stability and better outcomes.
Ever since the first anticancer drugs were discovered combination therapies were in the focus of attention because very few single drugs carry the potential to fundamentally change the trajectory of cancer. In oncology, nearly 80% of all clinical trials involve the study of combination therapies. However, the rational design of drug combinations remains a challenge due to the complex molecular networks that contribute to feedback mechanisms between drugs and the huge diversity between individual patients; tumors and the microenvironment in which they are embedded [2-3] .
To bridge this gap, we have developed Optim.AI™, a functional precision medicine platform that combines small data AI with biological experiments to analyze the functional response of drug-dose combinations. The foundation of Optim.AI™ was based on the discovery that responses of complex biological systems to therapeutic intervention can be represented by smooth parabolic surfaces, via second-order polynomial equations [4], where they can accurately identify globally optimal drug combination parameters [4-8]. The importance of first- and second-order terms in determining parabolic surfaces of therapeutic responses extends from cell cultures and preclinical models to humans and can be used to optimize drug-selection and -dosage in the clinical setting [9-12].
Conventional development of combination therapies relies on molecular properties and clinical outcome data from highly diverse patient populations to determine relative probabilities for individual patients. The high diversity between different tumors complicates approaches such as network modeling algorithms based on predetermined synergies. Optim.AI™ does not require this information to identify optimal drug combinations.
Optim.AI™ can identify optimal drug-dose combinations across different mechanisms of action regardless of the nature and stage of disease and has been applied at multiple stages in drug development [13-19] . The clinical validation of Optim.AI™ has been demonstrated across both hematological and solid cancers [13-19] .
Optim.AI™ is performed on fresh patient tissue that is sent to the KYAN laboratory in Singapore within 48 hours. Fresh frozen tissue or patient derived organoids can also be tested. For testing immunotherapies, matching blood mononuclear cells must be provided together with the tumor sample. The test results are usually sent to the ordering physician within ten days of the sample arriving at the KYAN laboratory in Singapore. The report lists most active 2-, 3-, and 4-drug combinations and indicates which are well established in the patient’s tumor type and for which data is available only from other tumor types or limited clinical studies.
Optim.AI™ has been approved as a commercial, laboratory-developed test by the Ministry of Health Singapore (MOH), the KYAN laboratories have been accredited by the College of American Pathologists (CAP) and certified according to ISO 13486.
The Optim.AI™ procedure has been validated in retrospective studies, prospective observational studies, prospective interventional studies, and several real-world evidence case series. In a study of patients with progressive non-Hodgkin lymphoma conducted by Goh et al., the test was successfully performed in 67 of 75 patients [16] . A total of 17 patients for whom no further
standard therapy was available were treated according to the test results, and 5 of these achieved a complete remission. Overall, progression-free survival was significantly better with therapy selected by the test than with therapy selected by the physician. Romidepsin and copanlisib-based dual combinations were found to be most effective.
Chan et al. used Optim.AI™ in patients with soft tissue sarcoma [13] . Positive and negative predictive values of 75% and 81% were found, the overall accuracy of the prediction was 77%, and of the 27 patients treated according to the test, 9 had stable disease and 7 had partial remission. It was also found that the combination of pazopanib with an investigational BET inhibitor performed best overall and was significantly superior to standard therapy. In breast cancer, Lim et al. found an accuracy of 80% for failure of the last previous therapy [19] . In a series of 51 patients with acute myeloid leukemia, 15 of 18 responsive patients and 10 of 11 non-responsive patients were correctly predicted, corresponding to positive and negative predictive values of 93% and 77%. Similarly, the response to the most used combination (azacytidine and venetoclax) could be determined with high accuracy (positive and negative predictive values of 100% and 86%).
In a recent publication by Lee et al., an update on the work in NHL, Optim.AI™ has
demonstrated predictive accuracy of 74.5%. OptimAI™-selected individualized therapy yielded an overall response rate of 59% and a hazard ratio for PFS of 0.56 in heavily pretreated patients compared with standard treatment (physician’s choice) [20] . Optim.AI™ has been used in over 20 different indications, and its average clinical accuracy is greater than 80%.
KYAN Technologies is a Singapore based biotechnology company using small data AI and biological experiments to redefine how therapies are developed and offered to patients. From preclinical drug development to personalized medicine, KYAN offers powerful solutions with its optimization of drug-dose combinations. KYAN has identified safe and effective therapies in various cancers.
A cross-functional team with decades of industry experience.
References