蜜桃传媒

Samudrala: CANDO simulates how thousands of compounds interact with the human body at once — like running millions of experiments in seconds. Instead of testing one drug at a time in a lab, CANDO is a powerful drug discovery platform based on foundation models of multiscale polypharmacology that helps scientists identify and design new medicines faster and more effectively via computing. Most drug discovery methods focus narrowly — one drug, one protein, one disease. CANDO looks at the big picture: how every drug interacts with every protein in the body and how these complex patterns relate to health and disease. This “systems-level” view helps us understand not just what works, but why it works and what else might work even better.

Samudrala: Traditional drug discovery can take over a decade to produce results. To overcome these limitations, we developed the multiscale CANDO paradigm, which was funded by an NIH Director’s Pioneer Award and then via NCATS ASPIRE Awards (among others). We wanted a smarter, faster way — so we created a platform that looks at how compounds affect the body as a whole, not just one protein. It was born out of the need for a system that leverages the power of computation to explore the vast universe of chemical and biological interactions.

Falls: CANDO was originally developed as a software package to predict new therapeutic uses for existing drugs. It takes a unique approach where it evaluates how a drug interacts multiple proteins that are involved in disease, which is in contrast to traditional, single-target approaches used by pharmaceutical companies. We have extended CANDO to evaluate other aspects of drug discovery such as to predict side effects and drug-drug interactions, as well as to design novel drugs with desired attributes.

Samudrala: AI is at the heart of CANDO. We use machine learning to analyze huge datasets of drug-protein interactions, predict new uses for existing drugs and design new ones with optimal properties. Think of AI in CANDO as the engine that sifts through the noise to find hidden patterns and connections that no human could easily spot. 

Falls: We have explored and integrated AI into multiple aspects of the platform. First, we expanded the multiscale understanding of CANDO by combining multiple data sources into a large, heterogeneous graph network and applying embedding techniques to extract the pertinent multiscale features for each drug in our library. We are also building more sophisticated models to predict drug-protein interactions to improve the data generation within the platform to improve the overall predictive power.

Samudrala: UB researchers in fields from biology and chemistry to data science and medicine can use CANDO to accelerate their work. Whether they are identifying potential treatments, designing new compounds or understanding disease mechanisms from single targets to lots of them — which is CANDO’s forte — they can obtain a jumpstart. It is available as a core service through the CTSI Informatics Core, with support for collaboration and integration into translational research.

Falls: I can see researchers using CANDO to design novel drug-like molecules with high potency against a newly identified protein target, predict potential side effects of a novel molecule that they synthesized or identify biomarkers implicated in disease they are studying. The applications are broad and we continue to find and develop new ways to apply this robust and flexible platform.

Samudrala: The future of CANDO is even more integrative and predictive. We are incorporating the latest advances in AI, structural biology — like AlphaFold3 — and other omics data to refine our models. We envision CANDO evolving into a universal platform not just for drug discovery, but also for biosensing, diagnostics and personalized medicine, thereby guiding decisions at every step of health care.

Falls: The evolution of CANDO will integrate patient level clinical and biological data. This will push CANDO toward precision medicine in a way that we will be able to tailor treatments or identify side effects that are unique to subgroups of people based on specific mutations, expression levels or other personalized features. The ability to assess if a newly designed drug would work better or worse for any given cohort of patients and identify potential side effects and toxicity prior to the drug entering clinical trials would save time, money and patient lives.

Both Samudrala and Falls note that CANDO has already helped identify new therapeutic possibilities, optimized compounds for better efficacy and safety, and reduced the time and cost of drug development. As a result, time from idea to experimental validation — in months instead of years — is potentially transformative in how quickly patients get new treatments.

Researchers interested in learning more about CANDO and its applications for their work can  through the CTSI Informatics Core service request portal.