One conversation is often all it takes to change the lives of children, siblings, spouses, friends, and parents. Something as simple as noticing a forgotten name or a repeated phrase could be enough for fear, sadness, and frustration to flood in. The symptoms spread into noticeable mood swings and disorientation. While memory loss is a typical aspect of aging, Alzheimer’s Disease (AD) is irreversible and far more debilitating, progressively affecting all aspects of daily life. It takes a toll not only on those affected, but also on their loved ones and caregivers. Despite being the most common neurodegenerative disease and affecting over 6.9 million Americans, there is no cure (Colović, 2013). There have been many attempts to develop treatments, but it has come to be known as a graveyard indication for drug development. With the emergence of two groundbreaking advancements in science and technology, molecular glues and generative AI, there is potential for a novel approach to addressing the disease’s pathology.
AD is a subtype of dementia characterized by amyloid beta plaque aggregation and neurofibrillary tangles that ultimately lead to brain atrophy. Amyloid beta 42 peptides (Aβ42) are more prone to forming deposits earlier in the onset of the disease due to aberrant cleavage of the amyloid precursor protein (Islam, 2025). If this early burden is not stopped or slowed, the progression will continue to worsen with increasing speed. Many previous attempts at mitigating symptoms and slowing the progression have fallen into three main categories: cholinesterase inhibitors, NMDA receptor antagonists, and amyloid targeting immunotherapies (National Institute on Aging, 2025).
Acetylcholinesterase is involved in breaking down acetylcholine, a key neurotransmitter, thereby decreasing neurotransmission (Colović, 2013). Drugs such as Aricept and Exelon decrease the enzyme’s activity to enhance stimulation of neurons and offset the pathology-induced suppression of neuronal activity (National Institute on Aging, 2025). The NMDA receptor is a ligand of glutamate, the primary excitatory neurotransmitter in the central nervous system, but excessive activation in AD leads to neurotoxicity, creating a positive feedback loop (Liu, 2019). NMDA receptor antagonists like memantine block stimulation to prevent neuron damage, focusing on treating the underlying cause rather than symptoms. While these have been somewhat effective, AD is very variable, and these drugs do not work for all patients. Molecular glues are a promising new approach to the third category of amyloid-targeting therapies.
Imagine the frustration of playing a claw machine, aiming to grab a prize and transfer it through the chute and out of the machine. No matter how hard you try, the prizes tend to clump together, and the claw ends up scraping the surface and coming up empty-handed. If you were to add glue to the claw that strengthens its hold on the prize as it picks it up, it would be far more successful in sending prizes through the chute. In this example, the machine itself is the brain, Aβ42 is the prize that clumps together, the claw is the E3 ligase, and the glue is the molecular glue that could increase the efficacy of the plaque clearance. The prize here is actually something that should be removed, and by adding a glue that increases its affinity to the tool that targets it for degradation, the removal is much more successful. The entire process of picking up and excreting a prize from the machine relates to ubiquitination, the process by which a target is brought to the proteasome, a machine inside cells that breaks the target down. The advantage of a molecular glue is that it is so small that it can diffuse through the blood-brain barrier, a lining of very tightly packed cells that work to prevent foreign bodies from invading the brain. In order for the target to get properly flagged, the glue must be precisely designed. For an aggregation-prone target like Aβ42, this design process is very difficult.
The gold standard for glue design is chemical screening, but this is very time-consuming. The Junction Tree Variational Autoencoder (JTVAE) is a generative artificial intelligence model that transforms molecules into hierarchical trees with maintained chemical rules and 3D visualization.2 This model can be used to screen for binding patterns by different glues between E3 ligases and Aβ42. Researchers have been able to identify which well-known E3 ligases form the strongest interactions between Aβ42 by screening different glues for blood-brain barrier permeability, intestinal absorption, hepatotoxicity, and other key components influencing efficacy.
To take a cause-based approach in developing a therapeutic for AD, the complexities of absorption into the brain and attack on pathology must be considered and creatively addressed. Molecular glues that can increase the interaction between Aβ42, an early-onset protein aggregate, and E3 ligase, a tagging enzyme that marks targets for degradation, have the potential to address those complexities. A tool like JTVAE that efficiently and exhaustively screens for molecular glue candidates offers a promising strategy for addressing this “graveyard” of drug development and restoring hope for the millions affected by AD.
Works Cited
- Colović, M. B., Krstić, D. Z., Lazarević-Pašti, T. D., Bondžić, A. M., & Vasić, V. M. (2013). Acetylcholinesterase inhibitors: pharmacology and toxicology. Current neuropharmacology, 11(3), 315–335. https://doi.org/10.2174/1570159X11311030006
- Islam, N. N., & Caulfield, T. R. (2025). Conditioned Generative Modeling of Molecular Glues: A Realistic AI Approach for Synthesizable Drug-like Molecules. Biomolecules, 15(6), 849. https://doi.org/10.3390/biom15060849
- Liu, J., Chang, L., Song, Y., Li, H., & Wu, Y. (2019). The Role of NMDA Receptors in Alzheimer's Disease. Frontiers in neuroscience, 13, 43. https://doi.org/10.3389/fnins.2019.00043
- How is alzheimer’s disease treated? | National Institute on Aging. (n.d.). https://www.nia.nih.gov/health/alzheimers-treatment/how-alzheimers-disease-treated