Transforming Neurotherapeutic Discovery With AI and Human Brain Organoids
- Brain diseases are the leading cause of disability and second-leading cause of death worldwide, yet most lack disease-modifying therapies because traditional models fail to predict clinical success.
- BrainStorm Therapeutics grows patient-derived 3D brain organoids (“mini-brains”) and trains multimodal AI on single-cell, imaging, and network-activity data to find targets conventional approaches miss.
- In a Parkinson’s proof-of-concept, GBA1-mutant midbrain organoids reproduced hallmark disease features and revealed dysregulated gene networks plus novel therapeutic targets.
- The “clinical trial in a dish” approach enables in silico drug-response simulation and patient stratification, and is now being applied to CDKL5 Deficiency Disorder.
- Introducing human relevance and AI early de-risks target selection in a field where clinical trial failure rates exceed 90%, cutting both time and cost.
Submitted by BrainStorm Therapeutics.
Originally published in Life Sciences Insights Magazine, August 2025
Brain diseases are the leading cause of disability and second-leading cause of death worldwide, a burden expected to double within 20 years. Yet despite decades of research, disease-modifying therapies remain out of reach for most neurological disorders due to the inability of traditional disease models to predict clinical success. At BrainStorm Therapeutics, we are leveraging artificial intelligence (AI) to tackle this challenge head-on. By combining advanced AI with human mini-brain models, we are accelerating the discovery of precision therapeutics for disorders such as Parkinson’s disease, Alzheimer’s disease, and epilepsy.
AI Meets Human Biology
Our platform takes a biology-first approach. We generate brain organoids: 3D “mini-brains” derived from patient stem cells. These organoids faithfully replicate disease-relevant cell types and neural circuits, offering an unprecedented window into human brain biology.
We use our brain organoids to generate single-cell RNA sequencing, high-content imaging, and functional network activity datasets. These rich biological datasets allow us to train multimodal AI models to map gene expression, track disease-related pathways, and uncover new therapeutic targets, many of which would be missed by conventional drug discovery approaches.
In our recent collaboration with NVIDIA, we adapted cutting-edge AI foundation models like scGPT and Geneformer to create high-resolution gene co-expression networks. This approach allows us to predict disease progression and identify dysregulated pathways associated with neurological disease. As highlighted in a recent NVIDIA blog, this AI-biological feedback loop enables faster, more precise experimental design and target discovery for validation in brain organoids.
Parkinson’s Disease as a Proof-of-Concept
We’ve applied our AI-powered platform to Parkinson’s disease (PD), the second most common neurodegenerative disorder. People with PD still rely on symptomatic treatments that haven’t changed in decades and offer only mild, temporary benefits that wane over time. Research progress has long been hindered by the inability of animal models to capture the disease’s human-specific complexity.
To overcome this, we developed midbrain organoids derived from patients with pathogenic GBA1 mutations, the most common genetic risk factor for PD. These organoids exhibit hallmark PD features, including dopaminergic neuron loss and impaired dopamine secretion. Analysis using our AI foundation model approach revealed dysregulated gene networks related to cell cycle, neurogenesis, and synaptic signaling, mirroring results from post-mortem brain tissue of idiopathic PD patients. The organoids also showed defects in lipid metabolism, demonstrating the ability of this model system to capture both generalizable and mutation-specific disease mechanisms. Our approach revealed potential therapeutic targets for PD, including several known GWAS hits as well as novel candidates.
Our drug discovery engine functions as a “clinical trial in a dish,” providing actionable insights into PD mechanisms and enabling in silico simulations of drug response. By integrating patient-specific biology with predictive AI, we aim to develop precision therapeutics that can prevent, halt, or reverse PD.
A Platform for Precision Neuroscience
Our platform enables us to:
Map disease-relevant gene networks and prioritize therapeutic targets
Perform virtual screening of candidate therapeutic targets
Identify shared and divergent mechanisms across brain diseases
Stratify patients by molecular subtype for more targeted clinical trials
Through an ongoing collaboration with the CURE5 Foundation, we’re currently applying this platform to discover new therapeutics for CDKL5 Deficiency Disorder, a rare genetic form of epilepsy. Our long-term goal is to build a multi-modal AI platform that bridges the gap between organoid biology and patient phenotypes.
Aligning with Industry Trends
The biopharma industry is undergoing a notable shift: biopharma M&A and licensing deals are increasingly focused on preclinical early-stage startups and assets. Recent transactions — such as Mitokinin/AbbVie, Caraway/Merck, and Aliada/AbbVie — reflect a growing appetite for preclinical-stage assets.
This trend is driven by both scientific opportunity and market necessity. With patent cliffs approaching and R&D productivity under pressure, pharmaceutical companies are increasingly turning to early, de-risked innovation to sustain their pipelines.
In this shifting landscape, platforms like ours that are rooted in human biology and AI are becoming increasingly important to how new drugs are discovered and developed.
Reducing Risk, Cost, and Time
Brain diseases remain one of the highest-risk areas in drug discovery, with clinical trial failure rates exceeding 90%. There are no disease-modifying therapies currently approved for PD or other common neurodegenerative diseases.
By introducing human relevance at the earliest stages of discovery and applying AI to de-risk target selection, we are reducing both the timeline and cost associated with early R&D, while increasing the odds of success in the clinic.
Get the latest thought leadership from California’s life sciences sector in the quarterly Life Sciences Insights magazine. Share your own news and insights with CLS Member Voice.
FAQ: AI and Human Brain Organoids in Neurotherapeutic Discovery
Brain diseases are the leading cause of disability and second-leading cause of death worldwide, yet most lack disease-modifying therapies. A key reason is that traditional disease models — especially animal models — fail to capture the human-specific complexity of the brain, so they poorly predict clinical success.
Brain organoids are 3D “mini-brains” grown from patient stem cells that replicate disease-relevant cell types and neural circuits. BrainStorm generates single-cell RNA sequencing, high-content imaging, and network-activity data from them to train multimodal AI models that map gene expression, track disease pathways, and uncover new therapeutic targets.
The team built midbrain organoids from patients with pathogenic GBA1 mutations — the most common genetic risk factor for PD. These reproduced hallmark features like dopaminergic neuron loss, and AI analysis revealed dysregulated gene networks mirroring post-mortem PD tissue, surfacing both known GWAS hits and novel therapeutic targets.
It describes BrainStorm’s drug discovery engine, which uses patient-derived organoids and predictive AI to model disease mechanisms and run in silico simulations of drug response. This provides actionable insights and helps design precision therapeutics aimed at preventing, halting, or reversing disease before human trials.
Brain diseases have clinical trial failure rates exceeding 90%. By introducing human-relevant biology at the earliest stages of discovery and using AI to de-risk target selection, BrainStorm reduces both the timeline and cost of early R&D while improving the odds of success in the clinic.