Symphony Predictions

Aria’s Symphony™ platform can, using interpretable disease-specific models, predict which drug candidates are likely to be efficacious. This allows our scientists to select and advance lead candidates that have the best likelihood of success.

We are constantly testing and improving Symphony™. One way we’ve always done that is by “rediscovering” support for successful clinical trials completed by other companies. Now we’re taking that a step further and sharing our predictions for Phase 2 candidates that have yet to report trial results.

The list below represents candidates that our Symphony™ drug discovery platform predicts will achieve efficacy in their respective Phase 2 trials. Stay tuned for additional quarterly updates.

clinical trials PREDICTIONS

DISEASE	DRUG	MECHANISM OF ACTION	CLINICAL TRIAL END DATE	CLINICAL TRIAL LINK	PREDICTION PUBLISHED DATE
Schizophrenia	Ulotaront	Trace Amine-Associated Receptor 1 & 5-hydroxytryptamine Type 1A Against	January 6, 2023	clinicaltrials.gov	December 23, 2022
Asthma	Tozorakimab	Anti-IL33 Antibody	February 6, 2023	clinicaltrials.gov	December 15, 2022
Psoriatic Arthritis	JNJ-77242113	IL-23 Receptor Antagonist	February 24, 2023	clinicaltrials.gov	December 15, 2022
Psoriasis	JNJ-77242113	IL-23 Receptor Antagonist	February 24, 2023	clinicaltrials.gov	December 15, 2022
Nonalcoholic Steatohepatitis (NASH)	HPG-1860	FXR Agonist	March 2023	clinicaltrials.gov	December 15, 2022
Amyotrophic Lateral Sclerosis (ALS)	Pridopidine	Sigma-1 Receptor Agonist & D2 Receptor Antagonist	March 2023	clinicaltrials.gov	December 15, 2022
Idiopathic Pulmonary Fibrosis (IPF)	BMS-986263	Serpin H1 Inhibitor	September 2022	clinicaltrials.gov	August 31, 2022
Atopic Dermatitis	CMK-389	Anti-IL18 Antibody	October 6, 2022	clinicaltrials.gov	August 31, 2022
Nonalcoholic Steatohepatitis (NASH)	Efinopegdutide	GLP-1/Glucagon Receptor Dual Agonist	October 20, 2022	clinicaltrials.gov	August 31, 2022
Type-2 Diabetes	Retatrutide	GLP-1/Glucagon Receptor/GIPR Tri-agonist	October 29, 2022	clinicaltrials.gov	October 25, 2022
Epilepsy	Darigabat	GABAA Positive Allosteric Modulator	November 2022	clinicaltrials.gov	August 31, 2022

SYMPHONY™

Aria’s drug discovery engine.

Drug discovery is complex. Our technology decodes complex biology in an unprecedented manner, combining the widest biomedical data landscapes available with proprietary, purpose-built artificial intelligence. Symphony™ is the only AI drug discovery platform that integrates and simultaneously analyzes unrelated heterogenous data in one process. We see the whole picture, increasing our chance of fully understanding the biology and detecting a signal others can’t see.

PREDICTION METHODOLOGY

Background

Aria’s proprietary drug-discovery AI-platform, Symphony™, is at the center of the company’s prediction capabilities. Using Symphony™ and its unprecedented ability to simultaneously integrate and interrogate the widest biomedical data landscapes, Aria decodes a disease’s complex biology by building an interpretable in silico model. These disease-specific models accurately predict which drug candidates will be significantly efficacious in the disease.

Validating Symphony™ predictions with “rediscovery”

One of the prime ways Aria’s researchers verify the viability of the disease-specific models built on Symphony™ is how well those models blindly “rediscover” previously investigated treatments. That is, how well the models do at retrospectively identifying drug candidates that succeeded in clinical trials even when those outcomes are hidden from Symphony™.

The way this works is that for a given disease, Aria researchers identify all treatments that have completed Phase 2 or beyond; all those drugs are then divided into two groups: training and testing. Each group is carefully chosen to ensure no overlap between drug targets, receptors, or ligands. This careful splitting is important to avoid data leakage, and thereby ensures all retrospective analyses that follow are fully blinded.

The training group of treatments is used to finalize the disease-specific models Symphony™ produces. Aria researchers can then examine the efficacy predictions Symphony™ makes for every treatment in the testing group. Because all treatments in the testing group are unlabeled to Symphony™, this is an excellent and unbiased way to measure the quality of predictions.

Aria has recently completed this process for more than 30 disease-specific models, allowing us to complete a fully blinded retrospective analysis on more than 400 completed Phase 2 trials. The result is that more than 80 percent of the treatments Symphony™ predicted to be efficacious passed their Phase 2.

The AI at work: Making prospective predictions

Retrospectively identifying clinical successes is one thing, but we understand that prospective predictions are more compelling. That’s why Aria is using those same 30+ disease models to share prospective predictions on ongoing Phase 2 treatments. This will enable us to continue to validate and further improve our prediction capabilities. The prospective predictions we provide publicly are only from the diseases we have recently modeled (30+ and counting) and for candidates that Symphony™ predicts as having a high likelihood of efficacy. In addition, similar to our own discovery efforts our researchers have examined the treatments and the trial designs for every prospective prediction. In some cases, we see rationale for efficacy, but have questions or concerns – those cases and our concerns are listed here.