RESUMEN
Opioids are small-molecule agonists of µ-opioid receptor (µOR), while reversal agents such as naloxone are antagonists of µOR. Here, we developed machine learning (ML) models to classify the intrinsic activities of ligands at the human µOR based on the SMILES strings and two-dimensional molecular descriptors. We first manually curated a database of 983 small molecules with measured Emax values at the human µOR. Analysis of the chemical space allowed identification of dominant scaffolds and structurally similar agonists and antagonists. Decision tree models and directed message passing neural networks (MPNNs) were then trained to classify agonistic and antagonistic ligands. The hold-out test AUCs (areas under the receiver operator curves) of the extra-tree (ET) and MPNN models are 91.5 ± 3.9% and 91.8 ± 4.4%, respectively. To overcome the challenge of a small data set, a student-teacher learning method called tritraining with disagreement was tested using an unlabeled data set comprised of 15,816 ligands of human, mouse, and rat µOR, κOR, and δOR. We found that the tritraining scheme was able to increase the hold-out AUC of MPNN models to as high as 95.7%. Our work demonstrates the feasibility of developing ML models to accurately predict the intrinsic activities of µOR ligands, even with limited data. We envisage potential applications of these models in evaluating uncharacterized substances for public safety risks and discovering new therapeutic agents to counteract opioid overdoses.
Asunto(s)
Aprendizaje Automático , Receptores Opioides mu , Humanos , Receptores Opioides mu/metabolismo , Receptores Opioides mu/agonistas , Ligandos , Animales , Analgésicos Opioides/farmacología , Ratones , Redes Neurales de la Computación , RatasRESUMEN
Biomolecules often exhibit complex free energy landscapes in which long-lived metastable states are separated by large energy barriers. Overcoming these barriers to robustly sample transitions between the metastable states with classical molecular dynamics (MD) simulations presents a challenge. To circumvent this issue, collective variable (CV)-based enhanced sampling MD approaches are often employed. Traditional CV selection relies on intuition and prior knowledge of the system. This approach introduces bias, which can lead to incomplete mechanistic insights. Thus, automated CV detection is desired to gain a deeper understanding of the system/process. Analysis of MD data with various machine-learning algorithms, such as principal component analysis (PCA), support vector machine, and linear discriminant analysis (LDA) based approaches have been implemented for automated CV detection. However, their performance has not been systematically evaluated on structurally and mechanistically complex biological systems. Here, we applied these methods to MD simulations of the MFSD2A (Major Facilitator Superfamily Domain 2A) lysolipid transporter in multiple functionally relevant metastable states with the goal of identifying optimal CVs that would structurally discriminate these states. Specific emphasis was on the automated detection and interpretive power of LDA-based CVs. We found that LDA methods, which included a novel gradient descent-based multiclass harmonic variant, termed GDHLDA, we developed here, outperform PCA in class separation, exhibiting remarkable consistency in extracting CVs critical for distinguishing metastable states. Furthermore, the identified CVs included features previously associated with conformational transitions in MFSD2A. Specifically, conformational shifts in transmembrane helix 7 and in residue Y294 on this helix emerged as critical features discriminating the metastable states in MFSD2A. This highlights the effectiveness of LDA-based approaches in automatically extracting from MD trajectories CVs of functional relevance that can be used to drive biased MD simulations to efficiently sample conformational transitions in the molecular system.
RESUMEN
Identifying optimal reaction coordinates for complex conformational changes and protein folding remains an outstanding challenge. This study combines collective variable (CV) discovery based on chemical intuition and machine learning with enhanced sampling to converge the folding free energy landscape of lasso peptides, a unique class of natural products with knot-like tertiary structures. This knotted scaffold imparts remarkable stability, making lasso peptides resistant to proteolytic degradation, thermal denaturation, and extreme pH conditions. Although their direct synthesis would enable therapeutic design, it has not yet been possible due to the improbable occurrence of spontaneous lasso folding. Thus, simulations characterizing the folding propensity are needed to identify strategies for increasing access to the lasso architecture by stabilizing the pre-lasso ensemble before isopeptide bond formation. Herein, harmonic linear discriminant analysis (HLDA) is combined with metadynamics-enhanced sampling to discover CVs capable of distinguishing the pre-lasso fold and converging the folding propensity. Intuitive CVs are compared to iterative rounds of HLDA to identify CVs that not only accomplish these goals for one lasso peptide but also seem to be transferable to others, establishing a protocol for the identification of folding reaction coordinates for lasso peptides.
Asunto(s)
Aprendizaje Automático , Péptidos , Pliegue de Proteína , Péptidos/química , Simulación de Dinámica Molecular , Termodinámica , Análisis DiscriminanteRESUMEN
Opioids are small-molecule agonists of µ-opioid receptor (µOR), while reversal agents such as naloxone are antagonists of µOR. Here we developed machine learning (ML) models to classify the intrinsic activities of ligands at the human µOR based on the SMILE strings and two-dimensional molecular descriptors. We first manually curated a database of 983 small molecules with measured E max values at the human µOR. Analysis of the chemical space allowed identification of dominant scaffolds and structurally similar agonists and antagonists. Decision tree models and directed message passing neural networks (MPNNs) were then trained to classify agonistic and antagonistic ligands. The hold-out test AUCs (areas under the receiver operator curves) of the extra-tree (ET) and MPNN models are 91.5±3.9% and 91.8± 4.4%, respectively. To overcome the challenge of small dataset, a student-teacher learning method called tri-training with disagreement was tested using an unlabeled dataset comprised of 15,816 ligands of human, mouse, or rat µOR, κOR, or δOR. We found that the tri-training scheme was able to increase the hold-out AUC of MPNN to as high as 95.7%. Our work demonstrates the feasibility of developing ML models to accurately predict the intrinsic activities of µOR ligands, even with limited data. We envisage potential applications of these models in evaluating uncharacterized substances for public safety risks and discovering new therapeutic agents to counteract opioid overdoses.
RESUMEN
Lasso peptides make up a class of natural products characterized by a threaded structure. Given their small size and stability, chemical synthesis would offer tremendous potential for the development of novel therapeutics. However, the accessibility of the pre-folded lasso architecture has limited this advance. To better understand the folding process de novo, simulations are used herein to characterize the folding propensity of microcin J25 (MccJ25), a lasso peptide known for its antimicrobial properties. New algorithms are developed to unambiguously distinguish threaded from nonthreaded precursors and determine handedness, a key feature in natural lasso peptides. We find that MccJ25 indeed forms right-handed pre-lassos, in contrast to past predictions but consistent with all natural lasso peptides. Additionally, the native pre-lasso structure is shown to be metastable prior to ring formation but to readily transition to entropically favored unfolded and nonthreaded structures, suggesting that de novo lasso folding is rare. However, by altering the ring forming residues and appending thiol and thioester functionalities, we are able to increase the stability of pre-lasso conformations. Furthermore, conditions leading to protonation of a histidine imidazole side chain further stabilize the modified pre-lasso ensemble. This work highlights the use of computational methods to characterize lasso folding and demonstrates that de novo access to lasso structures can be facilitated by optimizing sequence, unnatural modifications, and reaction conditions like pH.
Asunto(s)
Bacteriocinas , Péptidos , Conformación Proteica , Péptidos/química , Bacteriocinas/química , Antibacterianos/químicaRESUMEN
Molecular simulations are commonly used to understand the mechanism of membrane permeation of small molecules, particularly for biomedical and pharmaceutical applications. However, despite significant advances in computing power and algorithms, calculating an accurate permeation free energy profile remains elusive for many drug molecules because it can require identifying the rate-limiting degrees of freedom (i.e., appropriate reaction coordinates). To resolve this issue, researchers have developed machine learning approaches to identify slow system dynamics. In this work, we apply time-lagged independent component analysis (tICA), an unsupervised dimensionality reduction algorithm, to molecular dynamics simulations with well-tempered metadynamics to find the slowest collective degrees of freedom of the permeation process of trimethoprim through a multicomponent membrane. We show that tICA-metadynamics yields translational and orientational collective variables (CVs) that increase convergence efficiency â¼1.5 times. However, crossing the periodic boundary is shown to introduce artifacts in the translational CV that can be corrected by taking absolute values of molecular features. Additionally, we find that the convergence of the tICA CVs is reached with approximately five membrane crossings and that data reweighting is required to avoid deviations in the translational CV.
Asunto(s)
Simulación de Dinámica Molecular , Termodinámica , EntropíaRESUMEN
Molecular simulations are commonly used to understand the mechanism of membrane permeation of small molecules, particularly for biomedical and pharmaceutical applications. However, despite significant advances in computing power and algorithms, calculating an accurate permeation free energy profile remains elusive for many drug molecules because it can require identifying the rate-limiting degrees of freedom (i.e., appropriate reaction coordinates). To resolve this issue, researchers have developed machine learning approaches to identify slow system dynamics. In this work, we apply time-lagged independent component analysis (tICA), an unsupervised dimensionality reduction algorithm, to molecular dynamics simulations with well-tempered metadynamics to find the slowest collective degrees of freedom of the permeation process of trimethoprim through a multicomponent membrane. We show that tICA-metadynamics yields translational and orientational collective variables (CVs) that increase convergence efficiency â¼1.5 times. However, crossing the periodic boundary is shown to introduce artefacts in the translational CV that can be corrected by taking absolute values of molecular features. Additionally, we find that the convergence of the tICA CVs is reached with approximately five membrane crossings, and that data reweighting is required to avoid deviations in the translational CV.
RESUMEN
The human brain is a dynamic multiplex of information, both neural (neurotransmitter-to-neuron, involving 1.5×1015 action potentials per minute) and immunological (cytokine-to-microglia, providing continuous immune surveillance via 1.5×1010 immunocompetent cells). This conceptualization highlights the opportunity of exploiting "information" not only in the mechanistic understanding of brain pathology, but also as a potential therapeutic modality. Arising from its parallel yet interconnected proteopathic-immunopathic pathogeneses, Alzheimer's disease (AD) enables an exploration of the mechanistic and therapeutic contributions of information as a physical process central to brain disease progression. This review first considers the definition of information and its relevance to neurobiology and thermodynamics. Then we focus on the roles of information in AD using its two classical hallmarks. We assess the pathological contributions of ß-amyloid peptides to synaptic dysfunction and reconsider this as a source of noise that disrupts information transfer between presynaptic and postsynaptic neurons. Also, we treat the triggers that activate cytokine-microglial brain processes as information-rich three-dimensional patterns, including pathogen-associated molecular patterns and damage-associated molecular patterns. There are structural and functional similarities between neural and immunological information with both fundamentally contributing to brain anatomy and pathology in health and disease. Finally, the role of information as a therapeutic for AD is introduced, particularly cognitive reserve as a prophylactic protective factor and cognitive therapy as a therapeutic contributor to the comprehensive management of ongoing dementia.