Abstract Designing efficient synthetic routes for a target molecule remains major challenge in organic synthesis. Atom environments are ideal, stand-alone, chemically meaningful building blocks providing high-resolution molecular representation. Our approach mimics chemical reasoning, and predicts reactant candidates by learning the changes of atom associated with reaction. Through careful inspection candidates, we demonstrate as promising descriptors studying reaction route prediction...

Accurate prediction of the binding affinity a protein-ligand complex is essential for efficient and successful rational drug design. Therefore, many methods have been developed. In recent years, since deep learning technology has become powerful, it also implemented to predict affinity. this work, new neural network model that predicts structure Our using ensemble multiple independently trained networks consist channels 3-D convolutional layers. was 3772 complexes from refined set...

A bstract In this preprint, we investigated whether AlphaFold2, AF2, can predict protein-peptide complex structures only with sequence information. We modeled the of 203 complexes from PepBDB DB and 183 PepSet. The were modeling concatenated sequences receptors peptides via poly-glycine linker. found that for more than half test cases, AF2 predicted bound good accuracy, C α -RMSD a peptide < 3.0 Å. For about 40% an accuracy 2.0 Our benchmark results clearly show has great potential to be...

Tokenization is an important preprocessing step in natural language processing that may have a significant influence on prediction quality. This research showed the traditional SMILES tokenization has certain limitation results tokens failing to reflect true nature of molecules. To address this issue, we developed atom-in-SMILES scheme eliminates ambiguities generic tokens. Our multiple chemical translation and molecular property tasks demonstrate proper impact In terms accuracy token...

Recently, predicting proteins three-dimensional (3D) structure from its sequence information has made a significant progress due to the advances in computational techniques and growth of experimental structures. However, selecting good models structural model pool is an important challenging task protein prediction. In this study, we present first application random forest based quality assessment (RFMQA) rank using features knowledge-based potential energy terms. The method predicts...

We present a new computational approach for constant pH simulations in explicit solvent based on the combination of enveloping distribution sampling (EDS) and Hamiltonian replica exchange (HREX) methods. Unlike methods variable continuous charge models, our method is discrete protonation states. EDS generates hybrid different A smoothness parameter s used to control heights energy barriers hybrid-state landscape. small value facilitates state transitions by lowering barriers. Replica between...

Abstract Here, we introduce a new molecule optimization method, MolFinder, based on an efficient global algorithm, the conformational space annealing and SMILES representation. MolFinder finds diverse molecules with desired properties efficiently without any training large molecular database. Compared recently proposed reinforcement-learning-based algorithms, consistently outperforms in terms of both given target property generation set novel molecules. The efficiency demonstrates that...

Fluorescent molecules, fluorophores or dyes, play essential roles in bioimaging. Effective bioimaging requires with diverse colors and high quantum yields for better resolution. An computational component to design novel dye molecules is an accurate model that predicts the electronic properties of molecules. Here, we present statistical machines predict excitation energies associated oscillator strengths a given molecule using random forest algorithm. The are closely related emission...

Protonation states of ionizable protein residues modulate many essential biological processes. For correct modeling and understanding these processes, it is crucial to accurately determine their pKa values. Here, we present four tree-based machine learning models for prediction. The models, Random Forest, Extra Trees, eXtreme Gradient Boosting (XGBoost), Light Machine (LightGBM), were trained on three experimental PDB datasets, two which included a notable portion internal residues. We...

Abstract The simplified molecular-input line-entry system (SMILES) is the most prevalent molecular representation used in AI-based chemical applications. However, there are innate limitations associated with internal structure of SMILES representations. In this context, study exploits resolution and robustness unique representations, i.e., SELFIES (SELF-referencIng Embedded strings), reconstructed from a set structural fingerprints, which proposed herein as vital representational tools for...

Machine learning interatomic potentials (MLIPs) promise quantum-level accuracy at classical force field speeds, but their performance hinges on the quality and diversity of training data. An efficient fully automated approach to sample chemical reaction space without relying human intuition, addressing a critical gap in MLIP development is presented. The method combines speed tight-binding calculations with selective high-level refinement, generating diverse datasets that capture both...

Abstract Background Acute myeloid leukemia (AML) is a highly aggressive cancer with 5-year survival rate of less than 35%. It characterized by significant drug resistance and abnormal energy metabolism. Mitochondrial dynamics metabolism are crucial for AML cell survival. fusion protein optic atrophy (OPA)1 upregulated in patients adverse mutations correlates poor prognosis. Method This study investigated targeting OPA1 TMQ0153, tetrahydrobenzimidazole derivative, to disrupt mitochondrial as...

Ab initio protein structure prediction is a challenging problem that requires both an accurate energetic representation of and efficient conformational sampling method for successful modeling. In this article, we present ab which combines recently suggested novel way fragment assembly, dynamic assembly (DFA) space annealing (CSA) algorithm. DFA, model structures are scored by continuous functions constructed based on short- long-range structural restraint information from library. Here, DFA...

We propose a modularity optimization method, Mod-CSA, based on stochastic global algorithm, conformational space annealing (CSA). Our method outperforms simulated in terms of both efficiency and accuracy, finding higher partitions with less computational resources required. The high values found by our are than, or equal to, the largest previously reported. In addition, can be combined other heuristic methods, implemented parallel fashion, allowing it to applicable large graphs more than 10...

Close stacking of arginine residues are often observed in protein structures despite the highly repulsive nature close like-charged groups. Physical factors stabilizing guanidinium ions side-chains have been previously studied water and protein-like environments, hydration free energy has emphasized to be an important factor. However, how pairs stabilized real proteins not fully understood yet. In this paper, we show that more frequently found interior than expected from frequency unpaired...

With the rapid improvement of machine translation approaches, neural has started to play an important role in retrosynthesis planning, which finds reasonable synthetic pathways for a target molecule. Previous studies showed that utilizing sequence-to-sequence frameworks is promising approach tackle retrosynthetic planning problem. In this work, we recast problem as language using template-free model. The model trained end-to-end and fully data-driven fashion. Unlike previous models...

Abstract Protein–ligand docking techniques are one of the essential tools for structure‐based drug design. Two major components a successful program an efficient search method and accurate scoring function. In this work, new called LigDockCSA is developed by using powerful global optimization technique, conformational space annealing (CSA), function that combines AutoDock energy piecewise linear potential (PLP) torsion energy. It shown CSA can find lower binding poses than Lamarckian genetic...

In the template-based modeling (TBM) category of CASP10 experiment, we introduced a new protocol called protein system (PMS) to generate accurate structures in terms side-chains as well backbone trace. protocol, global optimization algorithm, conformational space annealing (CSA), is applied three layers TBM procedure: multiple sequence-structure alignment, 3D chain building, and side-chain re-modeling. For developed energy function which includes distance restraint Lorentzian type (derived...

We present a computational scheme to compute the pH-dependence of binding free energy with explicit solvent. Despite importance pH, effect pH has been generally neglected in calculations because lack accurate methods model it. To address this limitation, we use constant-pH methodology obtain true ensemble multiple protonation states titratable system at given and analyze using Bennett acceptance ratio (BAR) method. The constant method is based on combination enveloping distribution sampling...

We present a new method for enhanced sampling constant-pH simulations in explicit water based on two-dimensional (2D) replica exchange scheme. The is significant extension of our previously developed simulation method, which enveloping distribution (EDS) coupled with one-dimensional (1D) Hamiltonian (HREM). EDS constructs hybrid from multiple discrete end state Hamiltonians that, this case, represent different protonation states the system. ruggedness and heights Hamiltonian's energy...