The FDA has shown a willingness to approve black box AI diagnostics, citing the approval of Artera AI's pathology model for predicting response to androgen deprivation therapy in prostate cancer, even...

Noetic AI uses a large language model-based platform to scrape the semantic web and annotate investigational drugs to create a ranked list for its therapeutics team.

Avi Mahajan believes the primary bottleneck in drug development is not the lack of good preclinical assets, but rather determining how well these assets work in patients.

China is able to run clinical trials far more cheaply than other countries due to lower labor costs and a more streamlined regulatory system where drugs are treated as "innocent before proven guilty."

Research from Leash Bio found that AI models predicting small molecule binding affinity can be confounded by the identity of the chemist who synthesized the molecule, as chemists have stylistic simila...

Axiom is developing a model to predict the toxicity impact of a small molecule on hepatocyte cells in a dish, which could save months in preclinical settings.

Despite a decade-long narrative, there is no strong evidence that improvements in preclinical drug design have led to better outcomes in human trials.

Noetic AI uses the generative capacity of its models to simulate knocking out specific genes to identify "nudge drugs" that can sensitize a tumor to other treatments.

Biopharmaceutical companies are increasingly focusing on acquiring preclinical assets from China rather than developing their drugs from scratch.

Avi Mahajan believes that reinforcement learning from experimental results in biology is fundamentally limited by the long, iterative process required to get feedback and the lack of verifiable ground...

A McKinsey study found that AI-designed drugs have a 5% to 10% lower failure rate in clinical trials compared to traditionally designed drugs.

Noetic AI collects human tumor data and profiles it across four modalities: pathology, spatial proteomics, whole plex spatial transcriptomics, and exome sequencing.

Avi Mahajan believes China's rise in drug development stems from its evolution from a generics manufacturer to a CRO hub like Wuxi AppTec, which created an ecosystem where drug design and wet lab assa...

Avi Mahajan believes that while China is producing interesting work in bio-AI, much of it involves scaling up concepts originally developed in the UK and America, without a "radically new" breakthroug...

Noetic AI's strategy is based on the idea that a "human simulator of biology," trained on rich data from real human tumors, could significantly reduce the current 97% failure rate of oncology trials.

Noetic AI operates on the thesis that the biomarkers defining patient response to a drug may be "non-human legible" and require a black box AI model to identify.