# Combining Chemical Biology and Machine Learning to Generate Reproducible Amyloid Fibrils > **NIH NIH F31** · UNIVERSITY OF PENNSYLVANIA · 2024 · $45,304 ## Abstract Project Summary Aging related disorders, such as Parkinson's Disease with dementia (PDD), Lewy Body with Dementia, and other Alzheimer Disease and related dementias (ADRDs), are debilitating neurological disorders that affect more than 8 million people in the US alone. Amyloid deposits, characterized by their insolubility and disruptive protein clumps, pose a distinct challenge in structural biology and medicine due to their atypical structure compared to soluble proteins. In PDD, the protein alpha-synuclein (αS) misfolds and aggregates into self-templating fibrils. These fibrils consist of repeating beta-sheet units, capable of adopting various shapes and surface structures, known as fibril morphologies (FMs). Ensuring consistent FM for in vitro assays remains a formidable challenge across literature given the variability introduced by factors such as salt concentration, buffer selection, and additive choices to name a few. These assays are vital for the assessment of compound binding and cellular toxicity, yet no expedient assay has been created to rigorously determine FM. The University of Pennsylvania (UPenn), sponsor, and principal investigator of this research proposal are uniquely positioned to accomplish the goals outlined below. UPenn provides a robust infrastructure and extensive resources for groundbreaking research, the sponsor contributes deep expertise in chemistry and biophysics, vital for the project's scientific rigor, and the principal investigator, with extensive experience in the relevant areas of this research, is well- equipped to lead and ensure the successful execution of the project. The project proposed relies on the hypothesis that proteolytic cleavage rates of a fibril are determined by stability of the fibril fold and enzyme steric clashes resulting in a specific kinetic proteolytic profile (KPP) which can distinguish FMs. Current methods require laborious assays to pinpoint FM, and with FM often varying unpredictably under identical preparations, the need is clear: a streamlined and efficient assay to guarantee uniformity across samples and literature. Therefore, in Aim 1 introduces an assay that is designed to determine FM by KPP. Pinpointing the FM is foundational to reproducibility and enables Aim 2: a statistically rigorous high- throughput screen (HTS) for exploration into the diversity of possible FMs. The envisioned HTS utilizes site selective chemistry via a cysteine to introduce mutations to the monomer structure of αS which is envisioned to induce conformational change when fibrils are formed. It is anticipated that a large FM library will be generated within the first few iterations which will not only expand the collective knowledge of the FM manifold, or coverage of all possible FMs, but also enable the creation of consistent preparations with specific features. Due to the number of potential combinations, Machine learning (ML) will be employed to guide future iterations of this HTS. By addressing these ... ## Key facts - **NIH application ID:** 10996436 - **Project number:** 1F31AG090063-01 - **Recipient organization:** UNIVERSITY OF PENNSYLVANIA - **Principal Investigator:** Ryann Michael Perez - **Activity code:** F31 (R01, R21, SBIR, etc.) - **Funding institute:** NIH - **Fiscal year:** 2024 - **Award amount:** $45,304 - **Award type:** 1 - **Project period:** 2024-09-05 → 2025-09-04 ## Primary source NIH RePORTER: https://reporter.nih.gov/project-details/10996436 ## Citation > US National Institutes of Health, RePORTER application 10996436, Combining Chemical Biology and Machine Learning to Generate Reproducible Amyloid Fibrils (1F31AG090063-01). Retrieved via AI Analytics 2026-05-26 from https://api.ai-analytics.org/grant/nih/10996436. Licensed CC0. --- *[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*