AI Reveals How 3-Secretase Selects Alzheimer’s Targets

Munich-based researchers from LMU, TUM, and DZNE have elucidated key mechanisms governing 3-secretase, an enzyme implicated in Alzheimer’s disease, potentially opening avenues for targeted therapeutics. Employing a novel ‘Comparative Physicochemical Profiling’ technique combined with explainable artificial intelligence, the team identified substrate characteristics beyond amino acid sequencing, revealing a preference for proteins exhibiting conformational flexibility within their transmembrane domains. Published this year in Nature Communications, the study identified several previously unknown substrates involved in immune regulation and carcinogenesis, suggesting the approach could extend to decoding interactions for other proteases and receptors, and ultimately accelerate drug development targeting these pathways.

Enzyme Specificity Decoded

Researchers at LMU Munich, the Technical University of Munich, and the German Center for Neurodegenerative Diseases have elucidated the mechanism by which γ-secretase – a protease capable of cleaving over 150 membrane proteins – selects its substrates. Unlike many proteases which identify targets via specific amino acid sequences, γ-secretase recognition is governed by a more holistic assessment of substrate properties.

The research team developed Comparative Physicochemical Profiling (CPP), a novel technique enabling comparison of the physicochemical characteristics of known substrates against reference proteins, to identify defining patterns. This approach, coupled with explainable artificial intelligence (XAI), revealed that substrates possess a distinct physicochemical profile extending across the transmembrane domain and adjacent sequences. Crucially, substrates exhibit a propensity for extended conformations – deviating from typical helical structures – in proximity to the cleavage site, a finding corroborated by structural data from enzyme-substrate complexes.

Application of the CPP method also identified previously unknown γ-secretase substrates, including proteins involved in immune regulation and carcinogenesis. This expanded understanding of substrate specificity has implications for Alzheimer’s drug discovery, as γ-secretase’s role in amyloid precursor protein processing is central to the disease’s pathology. The authors suggest the CPP approach is broadly applicable, offering a framework for decoding the interplay of sequence, structure, and function in other proteases and receptors, potentially accelerating the development of therapeutically relevant compounds, including small-molecule drugs and antibodies, with improved specificity.

Physicochemical Profiling and AI Application

The study’s methodological innovation lies in the integration of quantitative physicochemical analysis with explainable artificial intelligence. While traditional approaches often focus on identifying specific amino acid motifs, this research demonstrates that γ-secretase substrate recognition is determined by a complex profile encompassing the entire transmembrane domain and adjacent regions. This holistic assessment, facilitated by Comparative Physicochemical Profiling (CPP), allows for the identification of subtle, yet critical, differences between substrates and non-substrates.

The application of CPP not only clarified the mechanism of γ-secretase action but also yielded novel insights into its substrate range. The identification of previously unknown substrates – including proteins involved in immune regulation and carcinogenesis – expands the known functional consequences of γ-secretase activity and necessitates a re-evaluation of its broader physiological role. These findings are particularly relevant to Alzheimer’s drug discovery, as a more comprehensive understanding of γ-secretase substrate specificity could inform the design of targeted therapies with reduced off-target effects.

The researchers emphasize the potential for extending the CPP methodology beyond γ-secretase. The framework established for decoding the interplay of sequence, structure, and function in proteases and receptors offers a powerful tool for investigating a wide range of biological processes. This adaptability could significantly accelerate the identification of therapeutically relevant compounds, including small-molecule drugs, peptides, or antibodies, designed to modulate protein-protein interactions or enzymatic activity with enhanced precision.

Therapeutic Implications and Future Research

The expanded substrate profile identified through this research offers new avenues for therapeutic intervention. Modulation of γ-secretase activity, particularly with compounds exhibiting enhanced specificity for particular substrates, may offer a more refined approach to treating Alzheimer’s disease and other conditions linked to aberrant γ-secretase processing. The identification of immune regulatory and carcinogenic proteins as substrates suggests potential implications beyond amyloid pathology, opening possibilities for novel therapeutic strategies targeting these pathways.

Furthermore, the CPP methodology itself represents a valuable asset in the broader context of drug development. The ability to predict substrate specificity, coupled with the transparency afforded by explainable AI, allows for in silico screening of potential therapeutic compounds, reducing the reliance on costly and time-consuming in vitro and in vivo studies. This accelerated screening process could significantly expedite Alzheimer’s drug discovery and the development of targeted therapies for a range of diseases.

The potential for applying CPP to other proteases and receptors extends beyond identifying novel drug targets. A deeper understanding of the interplay between sequence, structure, and function will be crucial for designing compounds that selectively modulate protein activity, minimising off-target effects and maximising therapeutic efficacy. This approach is particularly relevant for complex proteases, where traditional structure-based drug design has proven challenging. The resulting insights may facilitate the development of precision medicines tailored to individual patient profiles and disease mechanisms.