Predictive Oncology Successfully Develops Predictive Models Derived from Never-Before-Seen Compounds for Prevalent Cancer Indications Including Breast, Colon and Ovary

Rhea-AI Summary

Predictive Oncology (NASDAQ: POAI) has successfully developed predictive models from 21 unique compounds sourced from the Natural Products Discovery Core (NPDC) at the University of Michigan Life Sciences Institute. The company's AI-driven platform evaluated these compounds for tumor response across breast, colon, and ovary cancer types.

Key findings include:

• Three compounds showed stronger tumor response than the benchmark drug Doxorubicin across all tested tumor types
• A fourth compound demonstrated strong response in ovary and colon models
• Three additional compounds showed significant 'hit responses' across all tumor types
• The ML model achieved 73% prediction coverage while only requiring 7% of possible wet lab experiments, potentially saving up to two years of laboratory testing

The NPDC library, one of the largest pharmaceutically viable natural products collections in the United States, contains specimens from global biodiverse hotspots. Natural products have historically contributed to at least half of approved small-molecule drugs in the past three decades.

Positive

• Three compounds outperformed benchmark drug Doxorubicin across all tumor types
• ML platform achieved 73% prediction accuracy while reducing lab testing by 93%
• Potential to save up to two years in laboratory testing time
• Successfully tested compounds against multiple cancer types (breast, colon, ovary)
• Partnership with major research institution (University of Michigan) provides access to extensive compound library

Negative

• Only 1% of available NPDC library tested so far
• Results still require further investigation and validation
• Additional testing and development needed before commercialization

Insights

Biotechnology Research Analyst

Predictive Oncology's announcement represents a significant technological validation of their AI-driven drug discovery platform. By successfully developing predictive models for 21 compounds from the University of Michigan's Natural Products Discovery Core against breast, colon, and ovary cancer samples, they've demonstrated their platform's ability to accelerate the drug candidate selection process.

The efficiency metrics are particularly impressive: their machine learning platform made confident predictions covering 73% of experiments after measuring only 7% of possible wet lab tests, potentially eliminating up to two years of laboratory testing. In drug discovery, where time equates to substantial costs, this represents meaningful acceleration of the traditional discovery process.

Three compounds demonstrated tumor response stronger than Doxorubicin (a standard chemotherapy benchmark) across all tested tumor types, with additional compounds showing promising activity in specific cancer types. This suggests their AI platform is effectively identifying promising candidates from natural products, which historically account for approximately half of approved small-molecule drugs.

While these findings represent very early-stage discovery work that would require years of development before potential commercialization, the real value here is the platform validation. By successfully identifying promising compounds from just 1% of the available NPDC library, POAI demonstrates capability that could attract additional partnerships beyond this collaboration. The announced intention for future collaborations with the NPDC suggests an ongoing relationship that could yield additional candidates from their extensive natural products library.

Pharmaceutical Development Expert

This announcement highlights how Predictive Oncology is addressing two fundamental challenges in drug development: reducing screening time and improving candidate selection efficiency. Their AI platform's ability to predict 73% of experimental outcomes after conducting only 7% of wet lab tests represents a significant acceleration of the traditional hit-to-lead process.

The identification of three compounds outperforming Doxorubicin across multiple tumor types is noteworthy, as beating established therapies in initial screening represents a promising starting point. Natural products offer unique structural diversity that synthetic libraries often lack, making them valuable sources for novel mechanisms of action against cancer.

From a development perspective, it's important to recognize these are extremely early findings. These compounds would still require extensive optimization, ADME studies, toxicity screening, and preclinical validation before any potential clinical testing. However, the partnership with Michigan's Natural Products Discovery Core provides access to a pharmaceutically relevant library with global biodiversity representation, increasing the probability of identifying candidates with novel pharmacophores.

The efficiency demonstrated here - eliminating potentially years of laboratory testing - could significantly impact the economics of early drug discovery by allowing research organizations to allocate resources more effectively. Rather than testing every compound against every cancer type, this predictive approach prioritizes the most promising experiments, potentially increasing the overall productivity of discovery efforts. This collaboration model, leveraging academic natural product resources with commercial AI capabilities, represents an increasingly viable pathway for identifying new oncology leads in a field where innovation often comes from unexplored chemical space.

Company successfully developed predictive models derived from 21 unique compounds from the Natural Products Discovery Core at the University of Michigan

Tumor response models for novel compounds represent true drug discovery using Predictive's active machine learning platform

PITTSBURGH, March 25, 2025 (GLOBE NEWSWIRE) -- Predictive Oncology Inc. (NASDAQ: POAI), a leader in AI-driven drug discovery, announced today that it has successfully developed predictive models derived from 21 unique compounds from the Natural Products Discovery Core (NPDC) at the University of Michigan Life Sciences Institute.

Predictive Oncology, in partnership with the NPDC, recently evaluated 21 novel compounds using Predictive’s active machine learning platform. The platform is used to shorten the time necessary to select drug candidates, while increasing the probability of technical success using live-cell tumor samples from its extensive biobank of frozen specimens.

The U-M Natural Products Discovery Core is home to a best-in-class library, and among one of the largest pharmaceutically viable natural products libraries in the United States, with specimens collected from biodiverse hotspots around the globe including Asia-Pacific, the Middle East, South America, North America and the Antarctic.

Natural products are specialized molecules with diverse biological activities. At least half of the small-molecule drugs approved during the past three decades were derived from these products, underscoring their importance in drug discovery and the potential to patent and market these assets.

“Three compounds consistently demonstrated strong tumor drug response across all tumor types tested and demonstrated a stronger response than Doxorubicin, a benchmark compound, across tumor types,” said Dr. Arlette Uihlein, SVP of Translational Medicine and Drug Discovery at Predictive Oncology. “A fourth drug showed a strong response in the ovary and colon models and three additional compounds demonstrated the most ‘hit responses’ across all three tumor types.”

“The efforts of this program and Predictive Oncology’s platform along with these novel compounds is tangibly driving and supporting true drug discovery,” Dr. Uihlein concluded.

Three tumor types — breast, colon and ovary — were selected for testing with 21 NPDC compounds and a benchmark known anti-cancer drug. After only measuring 7% of the possible wet lab experiments, the predictive ML model was capable of making confident predictions to cover a total of 73% of all experiments, virtually eliminating up to two years of laboratory testing.

“Demonstrating that these natural compounds have such strong anti-tumor activity against several human tumor types strongly supports further investigations into these compounds and additional compounds, especially when considering that these results were achieved by including only about 1% of the available NPDC library,” added NPDC Director Dr. Ashu Tripathi. “As we review these first data sets, we look forward to future collaborations with Predictive Oncology to test more of the hundreds of compounds in our drug discovery pipeline, as well as publishing our results.”

About Predictive Oncology
Predictive Oncology is on the cutting edge of the rapidly growing use of artificial intelligence and machine learning to expedite early drug discovery and enable drug development for the benefit of cancer patients worldwide. The company’s scientifically validated AI platform, PEDAL, is able to predict with 92% accuracy if a tumor sample will respond to a certain drug compound, allowing for a more informed selection of drug/tumor type combinations for subsequent in-vitro testing. Together with the company’s vast biobank of more than 150,000 assay-capable heterogenous human tumor samples, Predictive Oncology offers its academic and industry partners one of the industry’s broadest AI-based drug discovery solutions, further complimented by its wholly owned CLIA laboratory facility. Predictive Oncology is headquartered in Pittsburgh, PA.

Investor Relations Contact:
Michael Moyer
LifeSci Advisors, LLC
mmoyer@lifesciadvisors.com

Forward-Looking Statements:
Certain matters discussed in this release contain forward-looking statements. These forward- looking statements reflect our current expectations and projections about future events and are subject to substantial risks, uncertainties and assumptions about our operations and the investments we make. All statements, other than statements of historical facts, included in this press release regarding our strategy, future operations, future financial position, future revenue and financial performance, projected costs, prospects, changes in management, plans and objectives of management are forward-looking statements. The words “anticipate,” “believe,” “estimate,” “expect,” “intend,” “may,” “plan,” “would,” “target” and similar expressions are intended to identify forward-looking statements, although not all forward-looking statements contain these identifying words. Our actual future performance may materially differ from that contemplated by the forward-looking statements as a result of a variety of factors including, among other things, factors discussed under the heading “Risk Factors” in our filings with the SEC. Except as expressly required by law, the company disclaims any intent or obligation to update these forward-looking statements.

FAQ

What breakthrough did Predictive Oncology (POAI) achieve in cancer drug discovery?

POAI developed predictive models from 21 NPDC compounds, with three compounds showing stronger tumor response than Doxorubicin across breast, colon, and ovary cancer types.

How much time could POAI's machine learning platform save in drug testing?

The platform could eliminate up to two years of laboratory testing by achieving 73% prediction coverage with only 7% of wet lab experiments.

Which cancer types were tested in POAI's recent compound evaluation?

The study tested compounds against three cancer types: breast, colon, and ovary.

How many compounds from NPDC showed strong anti-tumor activity in POAI's testing?

Three compounds showed strong response across all tumor types, one compound showed strong response in ovary and colon, and three additional compounds showed significant hit responses.

What percentage of the NPDC library was used in POAI's successful drug discovery study?

Only about 1% of the available NPDC library was used in this successful study.