ISSN: 2940-3243

Commentary

AlphaFold 2: Insights and Future Directions in Psychiatry

by Pascal Büttiker¹, Simon Weissenberger^1,2 Martin Anders¹, Jiri Raboch¹, George B. Stefano¹ and Amira Boukherissa^3,4

Abstract

The intricate links between protein misfolding and psychiatric disorders have drawn increasing attention in recent years. Several published studies have highlighted the contributions of misfolded proteins to the pathophysiology of conditions that include schizophrenia, depression, Alzheimer’s disease, and prion diseases. Recent breakthroughs in computational approaches, exemplified by the artificial intelligence application AlphaFold 2, have enabled us to predict protein structures with remarkable precision, resulting in unprecedented insights into the molecular mechanisms driving psychiatric illnesses. However, challenges remain in the application of computational power and its accuracy, particularly given the unrealized potential of quantum computing (QC). This commentary examines current advances in protein-folding and its impact on psychiatric research, addresses several of the ongoing challenges, and discusses the future promise of QC for deepening our understanding and advancing the treatment of neuropsychiatric disorders.

Keywords: bioinformatics, AlphaFold2 (AF2), psychiatry, quantum computing (QC)

Proteins play critical roles in normal cellular functions and maintain specific structures that define their individual roles in biological processes. Misfolded proteins have been linked to several well-known psychiatric and neurodegenerative disorders. To address these concerns, we need to improve our collective understanding of both normal and dysregulated protein folding (Ciechanover & Kwon, 2015; Stefano, Büttiker, Weissenberger, Anders, et al., 2023). Traditional methods used to elucidate protein structure such as X-ray crystallography and nuclear magnetic resonance are time-consuming and costly. The AlphaFold 2 (AF2) artificial intelligence (AI) application developed by DeepMind (Alphabet) has revolutionized the prediction of highly accurate protein structures, thereby accelerating disease research (Borkakoti & Thornton, 2023; Büttiker et al., 2024; Jumper et al., 2021; McDonough, 2024). Despite these advantages, AF2 is not fully capable of capturing complex folding patterns and interactions within the dynamic environment of the brain. Emerging quantum computing (QC) methods offer potential solutions with faster calculations and advanced modeling capabilities (Stefano, 2024).

The Impact of AF2 on Psychiatric Research

AF2 has revolutionized computational biology and has provided accurate predictions of thousands of protein structures, making protein structure determinations more accessible to the medical research community (Varadi et al., 2022). Researchers have applied AF2 to elucidate the structures of proteins involved in neurotransmission, synaptic plasticity, and neuronal development, all critical pathways implicated in psychiatric conditions, notably schizophrenia and bipolar disorder (Büttiker et al., 2024; Stefano, Büttiker, Weissenberger, Esch, et al., 2023). We surmise that the new-found capacity to model proteins previously designated as difficult to characterize will help researchers identify novel therapeutic targets for these disorders. AF2-based strategies may also be used to identify targets that may respond to lifestyle modifications that can be addressed as a means to prevent a given psychiatric manifestation, e.g., targeting neurobiological systems associated with motivation and reward (Michaelsen & Esch, 2023). Despite these remarkable achievements, several limitations remain. While AF2 is highly effective at predicting static structures, structural analyses of complex dynamic protein-protein interactions and post-translational modifications remain challenging (Gomes et al., 2022). The AF2 program in its current form does not intrinsically account for the cellular environmental factors (pH, temperature, ion concentrations, and/or micro- and macro-environmental thermodynamics) that critically influence protein folding in the brain (Guo et al., 2022). Many proteins involved in neuronal signaling and that have been implicated in one or more psychiatric disorders undergo constant conformational changes. These changes are challenging to predict using current computational models.

Protein Folding Challenge

It is important to recognize that protein folding is a quantum mechanical process, involving atomic interactions that classical and current computing strategies cannot yet simulate efficiently. Based on fundamental units known as qubits and the capacity to leverage superposition and entanglement, QC can transform our approach to this problem by processing vast numbers of configurations simultaneously, thereby providing more accurate simulations of protein folding pathways (Stefano, 2023, 2024).