Truly differentiated therapeutics from a biology-first AI platform.
Auransa is a clinical-stage, AI-native biopharmaceutical company based in Palo Alto, California. It builds cancer therapies the way most drug companies do not - by studying the messy, heterogeneous biology of human disease first, and only then predicting which compounds might work, and for which patients.
Most drug programs start with a target: a single protein or pathway a molecule is designed to hit. That choice happens early, and every later decision inherits its assumptions. Auransa's approach inverts the order. Its proprietary platform, the SMarTR Engine, mines large public and proprietary datasets - gene-expression profiles across thousands of tumors - to map the molecular subtypes hidden inside a single named disease. Only after that map exists does the engine predict compounds, and it pairs each prediction with the patient subtype most likely to respond.
The logic is simple to state and hard to execute: cancers that share a name do not share a biology. A "liver cancer" is really a collection of molecular diseases. By embracing that heterogeneity instead of averaging it away, Auransa aims to raise the odds that a drug reaches the clinic - and reaches the right people once it gets there.
"At Auransa, AI is first used to make sense of complex, heterogeneous human disease biology - before working on molecules."
Machine learning and advanced analytics digest molecular data from thousands of tumors to surface distinct disease subtypes - the structure inside the noise.
The engine searches approved, investigational and natural compounds at once, predicting which ones match each subtype. About half of predicted compounds show development potential after initial testing.
Each candidate is tied to the patient phenotypes most likely to respond, so the "right patient" question is answered before a trial begins - not after.
Note: hit-rate and process details are drawn from public interviews and company materials and are approximate.
An AI-derived RNA transcription modulator for advanced liver cancer (hepatocellular carcinoma) that affects the expression of genes cancer cells rely on for growth. Early data has shown a partial response and stable disease in pretreated patients, with no dose-limiting toxicities reported to date.
Phase 1 · US & AsiaA cardioprotective agent designed to reduce the heart toxicity of the workhorse chemotherapy doxorubicin - while aiming to increase its anti-cancer activity. Advancing under a partnership with Lee's Pharmaceutical.
Heart-safe chemoAdditional preclinical programs derived from the SMarTR Engine, spanning cancer subtypes and neglected diseases identified through the platform's target-free discovery process.
PreclinicalThe core AI/ML platform. It has been turned on problems beyond oncology, too - in a COVID-19 project, three of five compounds it flagged showed promising results alongside remdesivir in high-containment testing.
ProprietaryAuransa sits in a crowded and fast-moving field. Companies like Recursion, Insitro, Exscientia, BenevolentAI and Tempus have all staked claims on applying machine learning to drug discovery. Many begin from molecular structure, imaging, or a chosen target and optimize outward.
Auransa's wedge is sequence: biology first, chemistry second, patient built in from the start. Rather than asking "what molecule best hits this target," it asks "what does this disease actually look like across patients, and what already-existing or novel compound fits each version of it." That framing makes drug repurposing a natural output - an approved but shelved molecule may simply have never met the right patient subtype.
Who it serves. Auransa is a B2B drug developer. Its direct relationships are with pharmaceutical partners such as Lee's Pharmaceutical, academic centers like the University of Southern California, and technology collaborators including POLARISqb. The ultimate beneficiaries are cancer patients with few remaining options.
The problem it attacks. Drug development is brutally expensive and most candidates fail - often because a molecule that could help someone reached the wrong population. By matching compounds to responders up front, Auransa is trying to move the odds before a single patient is enrolled.
No predefined target. The disease data defines the direction.
Responder subtypes identified before trials, not after failures.
Approved, investigational and natural molecules considered together.
Auransa has raised roughly $34.2M in total, anchored by a $15.7M Series A that closed in 2020. It operates as an AI-native drug developer - discovering candidates in-house, advancing them into trials, and partnering with pharma companies to fund and scale clinical development.
Collaboration to advance clinical development of AU-018 (heart-safe chemotherapy) and AU-409 for advanced liver cancers.
Research collaboration pairing Auransa's AI with quantum-computing molecular design to find treatments for triple-negative breast cancer and neglected women's diseases.
Clinical collaboration supporting the Phase 1 trial of AU-409 in liver cancer.
Pek Lum and Viwat Visuthikraisee start the company to apply a biology-first AI approach to drug discovery.
Auransa presents new preclinical results on its AI-derived hepatocellular carcinoma candidate.
First patient dosed in the AU-409 liver cancer trial, and the company closes a $15.7M Series A.
Auransa partners with POLARISqb to combine AI and quantum computing for neglected women's diseases.
The collaboration reports identifying promising treatment targets for triple-negative breast cancer.
Interviews & talks: search "Pek Lum Auransa" on YouTube for founder interviews on machine learning and drug discovery. Platform overview: Sociable feature on the SMarTR Engine.