Posted on

Chasing Alzheimer’s with Blood and Artificial Intelligence

Will Atkinson | November 2nd, 2023

Pull open your drawer, unfasten the lid of a pill bottle, and look inside to discover decades of labor and billions of dollars of resources dedicated to alleviating your headache. Between the two inextricably linked fields of drug discovery and development, it has been estimated that the annual nationwide expenses in the field total to roughly $139.2 billion. Alongside other elite institutions, “Vanderbilt ranks as a leader in [pharmacology] for academia, industry and government,” performing research to identify new therapeutic targets and mechanisms to remedy formidable diseases. Among many other brilliant scientists, third year pharmacology PhD student Emma Webb leads the charge with her work on the PAR4 receptor in the Hamm Lab , hoping to develop an improved anti-inflammatory medication.

PAR4 (protease-activated receptor 4) is part of the PAR family, a family of GPCR receptors that is found on the surface of platelet cells. These receptors are activated by enzymes called serine proteases that cleave tethered ligands. These ligands go on to function as agonists, which initiate several pathways that ultimately serve the function of activating the platelets’ coagulation cascade. 

PAR4 Activation Pathway

In her breakthrough 2017 paper, Dr. Heidi Hamm, PI of the Hamm Lab, discovered the discrepancy between PAR1 and PAR4, which formed the basis for the PAR4 project. Hamm discerned that PAR1 requires a low level of the enzyme thrombin to be activated and, consequently, propagates a nearly constant – albeit low-level – platelet activating pathway; conversely, PAR4 necessitates a higher concentration of thrombin in order to be activated, and elicits a far more robust coagulation pathway. Moreover, this high level of PAR4-induced platelet aggregation occasionally manifests as pathological: its sometimes constitutive activation and the resulting hypercoagulation have been linked to Alzheimer’s disease, acute kidney disease, and many other conditions.

Photo of Emma Webb

It was not, however, intrinsic interest in PAR4, or even the heroic goal of curing Alzheimer’s disease, that provoked Webb’s interest in the project. Rather, since the inception of her PhD, it has been the uniquely comprehensive “early stage drug discovery team between the Lindsley, Meiler, and Hamm labs” (which focus on medicinal chemistry, computational modeling, and pharmacology, respectively) that spurs Webb’s interest in PAR4 – with the three labs converging on the receptor project. Within this massive undertaking, Webb’s niche identifies the short-downstream effects of the activation of PAR4 and how they directly trigger platelet aggregation. Unfortunately, the structure of the PAR4 binding pocket is currently unknown, obfuscating Webb’s ability to study these PAR4 pathways.

Webb and the rest of the drug discovery triumvirate employ an innovative technique known as homology modeling in order to circumvent this issue. The purpose of a homology model is to analyze the structure and the quality of ligand docking on structurally understood receptors in order to infer specific unknown attributes about a similar receptor. 

Homology Modeling Workflow

These homology models make use of artificial intelligence’s ability to detect patterns between the primary structure and the resulting shape of hundreds of thousands of known proteins. In the case of PAR4, the primary structure is known, so the homology model – which makes the assumption that it features a binding pocket that closely resembles that of PAR1 – seeks to predict how the tethered ligand activates the receptor, and the shape of both of those subunits. Further, “the homology model [generates] different versions of the receptor [before] doing more computational refine[ment] to find the…most probable structure,” based on overall predicted stability. 

One critical fallback of a homology model is that the computationally generated structure is entirely subject to error on the part of the model. These models are generally accurate, but several assumptions are made during the process of the model building. In order to mitigate the downside of the assumptions that are made, a new homology model called AlphaFold was released in 2021, which provides scientists with the ability to generate 3D structures that, “by making use of x-ray crystal structure libraries, does not require the user to input known structures,” unmooring us from our preconceived notions about these receptors. With these theoretical protein structures, Webb uses a model to “run a binding simulation between the ligand and the receptor.” In order to test the validity of these modeled ligand-receptor interactions, Webb and co. run experiments with the agonists – generated by the Lindsley Lab –  that the model deems most probable, and test their resulting activation on platelets using flow cytometry.

If this three-headed group is able to find the shape of the binding pocket and ascertain specific details about the downstream pathways, they “will likely file for a patent and partner with a pharmaceutical company” to develop an anti-inflammatory drug [to downregulate PAR4 function] that could “be the new ibuprofen.” While it may be “too soon to know” exactly where the PAR4 project will go, Webb and Vanderbilt’s Pharmacology team is bounding towards the future.

Bibliography

  1. Clinical Pharmacology | Department of Medicine. [accessed 2023 Oct 15]. https://medicine.vumc.org/divisions/clinical-pharmacology
  2. Emma Webb | Department of Pharmacology | Vanderbilt University. [accessed 2023 Oct 15]. https://medschool.vanderbilt.edu/pharmacology/person/emma-webb/
  3. Hameduh T, Haddad Y, Adam V, Heger Z. Homology modeling in the time of collective and 
  4. artificial intelligence. Computational and Structural Biotechnology Journal. 2020;18:3494–3506. doi:10.1016/j.csbj.2020.11.007
  5. Highly accurate protein structure prediction with AlphaFold | Nature. [accessed 2023 Oct 15]. https://www.nature.com/articles/s41586-021-03819-2
  6. Home. Vanderbilt University. [accessed 2023 Oct 15]. https://lab.vanderbilt.edu/hamm-lab/
  7. Homology Modeling – an overview | ScienceDirect Topics. [accessed 2023 Oct 15]. https://www.sciencedirect.com/topics/biochemistry-genetics-and-molecular-biology/homology-modeling
  8. Homology Modeling- Definition, Steps, Diagram, Uses. [accessed 2023 Oct 15]. https://microbenotes.com/homology-modeling/
  9. Lindsley Lab | Vanderbilt University. [accessed 2023 Oct 15]. https://lab.vanderbilt.edu/lindsleylab/
  10. Rigg RA, Healy LD, Chu TT, Ngo ATP, Mitrugno A, Zilberman-Rudenko J, Aslan JE, Hinds MT, Vecchiarelli LD, Morgan TK, et al. Protease-activated receptor 4 (PAR4) activity promotes platelet granule release and platelet-leukocyte interactions. Platelets. 2019 [accessed 2023 Oct 15];30(1):126–135. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6397092/. doi:10.1080/09537104.2017.1406076
  11. The Unbearable Cost of Drug Development: Deloitte Report Shows 15% Jump in R&D to $2.3 Billion. [accessed 2023 Oct 15]. https://www.genengnews.com/gen-edge/the-unbearable-cost-of-drug-development-deloitte-report-shows-15-jump-in-rd-to-2-3-billion/
  12. Thrombin – an overview | ScienceDirect Topics. [accessed 2023 Oct 15]. https://www.sciencedirect.com/topics/medicine-and-dentistry/thrombin
  13. Wilson SJ, Ismat FA, Wang Z, Cerra M, Narayan H, Raftis J, Gray TJ, Connell S, Garonzik S, Ma X, et al. PAR4 (Protease-Activated Receptor 4) Antagonism With BMS-986120 Inhibits Human Ex Vivo Thrombus Formation. Arteriosclerosis, Thrombosis, and Vascular Biology. 2018 [accessed 2023 Oct 15];38(2):448–456. https://www.ahajournals.org/doi/10.1161/ATVBAHA.117.310104. doi:10.1161/ATVBAHA.117.310104
By Will Atkinson