AbLEF: antibody language ensemble fusion for thermodynamically empowered property predictions.

MOTIVATION: Pre-trained protein language and/or structural models are often fine-tuned on drug development properties (i.e. developability properties) to accelerate drug discovery initiatives. However, these models generally rely on a single structural conformation and/or a single sequence as a molecular representation. We present a physics-based model, whereby 3D conformational ensemble representations are fused by a transformer-based architecture and concatenated to a language representation to predict antibody protein properties. Antibody language ensemble fusion enables the direct infusion of thermodynamic information into latent space and this enhances property prediction by explicitly infusing dynamic molecular behavior that occurs during experimental measurement. RESULTS: We showcase the antibody language ensemble fusion model on two developability properties: hydrophobic interaction chromatography retention time and temperature of aggregation (Tagg). We find that (i) 3D conformational ensembles that are generated from molecular simulation can further improve antibody property prediction for small datasets, (ii) the performance benefit from 3D conformational ensembles matches shallow machine learning methods in the small data regime, and (iii) fine-tuned large protein language models can match smaller antibody-specific language models at predicting antibody properties. AVAILABILITY AND IMPLEMENTATION: AbLEF codebase is available at https://github.com/merck/AbLEF.

AbMelt: Learning antibody thermostability from molecular dynamics.

Antibody thermostability is challenging to predict from sequence and/or structure. This difficulty is likely due to the absence of direct entropic information. Herein, we present AbMelt where we model the inherent flexibility of homologous antibody structures using molecular dynamics simulations at three temperatures and learn the relevant descriptors to predict the temperatures of aggregation (Tagg), melt onset (Tm,on), and melt (Tm). We observed that the radius of gyration deviation of the complementarity determining regions at 400 K is the highest Pearson correlated descriptor with aggregation temperature (rp = -0.68 ± 0.23) and the deviation of internal molecular contacts at 350 K is the highest correlated descriptor with both Tm,on (rp = -0.74 ± 0.04) as well as Tm (rp = -0.69 ± 0.03). Moreover, after descriptor selection and machine learning regression, we predict on a held-out test set containing both internal and public data and achieve robust performance for all endpoints compared with baseline models (Tagg R2 = 0.57 ± 0.11, Tm,on R2 = 0.56 ± 0.01, and Tm R2 = 0.60 ± 0.06). In addition, the robustness of the AbMelt molecular dynamics methodology is demonstrated by only training on <5% of the data and outperforming more traditional machine learning models trained on the entire data set of more than 500 internal antibodies. Users can predict thermostability measurements for antibody variable fragments by collecting descriptors and using AbMelt, which has been made available.

Mechanistic insights on novel small molecule allosteric activators of cGMP-dependent protein kinase PKG1α.

cGMP-dependent protein kinase (PKG) represents a compelling drug target for treatment of cardiovascular diseases. PKG1 is the major effector of beneficial cGMP signaling which is involved in smooth muscle relaxation and vascular tone, inhibition of platelet aggregation and signaling that leads to cardioprotection. In this study, a novel piperidine series of activators previously identified from an ultrahigh-throughput screen were validated to directly bind partially activated PKG1α and subsequently enhance its kinase activity in a concentration-dependent manner. Compounds from initial optimization efforts showed an ability to activate PKG1α independent of the endogenous activator, cGMP. We demonstrate these small molecule activators mimic the effect of cGMP on the kinetic parameters of PKG1α by positively modulating the KM of the peptide substrate and negatively modulating the apparent KM for ATP with increase in catalytic efficiency, kcat. In addition, these compounds also allosterically modulate the binding affinity of cGMP for PKG1α by increasing the affinity of cGMP for the high-affinity binding site (CNB-A) and decreasing the affinity of cGMP for the low-affinity binding site (CNB-B). We show the mode of action of these activators involves binding to an allosteric site within the regulatory domain, near the CNB-B binding site. To the best of our knowledge, these are the first reported non-cGMP mimetic small molecules shown to directly activate PKG1α. Insights into the mechanism of action of these compounds will enable future development of cardioprotective compounds that function through novel modes of action for the treatment of cardiovascular diseases.

MolPROP: Molecular Property prediction with multimodal language and graph fusion.

Pretrained deep learning models self-supervised on large datasets of language, image, and graph representations are often fine-tuned on downstream tasks and have demonstrated remarkable adaptability in a variety of applications including chatbots, autonomous driving, and protein folding. Additional research aims to improve performance on downstream tasks by fusing high dimensional data representations across multiple modalities. In this work, we explore a novel fusion of a pretrained language model, ChemBERTa-2, with graph neural networks for the task of molecular property prediction. We benchmark the MolPROP suite of models on seven scaffold split MoleculeNet datasets and compare with state-of-the-art architectures. We find that (1) multimodal property prediction for small molecules can match or significantly outperform modern architectures on hydration free energy (FreeSolv), experimental water solubility (ESOL), lipophilicity (Lipo), and clinical toxicity tasks (ClinTox), (2) the MolPROP multimodal fusion is predominantly beneficial on regression tasks, (3) the ChemBERTa-2 masked language model pretraining task (MLM) outperformed multitask regression pretraining task (MTR) when fused with graph neural networks for multimodal property prediction, and (4) despite improvements from multimodal fusion on regression tasks MolPROP significantly underperforms on some classification tasks. MolPROP has been made available at https://github.com/merck/MolPROP . SCIENTIFIC CONTRIBUTION: This work explores a novel multimodal fusion of learned language and graph representations of small molecules for the supervised task of molecular property prediction. The MolPROP suite of models demonstrates that language and graph fusion can significantly outperform modern architectures on several regression prediction tasks and also provides the opportunity to explore alternative fusion strategies on classification tasks for multimodal molecular property prediction.

Structural insights into selective small molecule activation of PKG1α.

cGMP-dependent protein kinase I-α (PKG1α) is a target for pulmonary arterial hypertension due to its role in the regulation of smooth muscle function. While most work has focused on regulation of cGMP turnover, we recently described several small molecule tool compounds which were capable of activating PKG1α via a cGMP independent pathway. Selected molecules were crystallized in the presence of PKG1α and were found to bind to an allosteric site proximal to the low-affinity nucleotide binding domain. These molecules act to displace the switch helix and cause activation of PKG1α representing a new mechanism for the activation and control of this critical therapeutic path. The described structures are vital to understanding the function and control of this key regulatory pathway.

Optimization and Mechanistic Investigations of Novel Allosteric Activators of PKG1α.

Activation of PKG1α is a compelling strategy for the treatment of cardiovascular diseases. As the main effector of cyclic guanosine monophosphate (cGMP), activation of PKG1α induces smooth muscle relaxation in blood vessels, lowers pulmonary blood pressure, prevents platelet aggregation, and protects against cardiac stress. The development of activators has been mostly limited to cGMP mimetics and synthetic peptides. Described herein is the optimization of a piperidine series of small molecules to yield activators that demonstrate in vitro phosphorylation of vasodilator-stimulated phosphoprotein as well as antiproliferative effects in human pulmonary arterial smooth muscle cells. Hydrogen/deuterium exchange mass spectrometry experiments with the small molecule activators revealed a mechanism of action consistent with cGMP-induced activation, and an X-ray co-crystal structure with a construct encompassing the regulatory domains illustrated a binding mode in an allosteric pocket proximal to the low-affinity cyclic nucleotide-binding domain.

Gamma radiation-induced preparation of chitosan-acrylic acid-1-vinyl-2-vinylpyrrolidone/multiwalled carbon nanotubes composite for removal of 152+154Eu, 60Co and 134Cs radionuclides.

Removal behaviors of 152+154Eu, 60Co, and 134Cs radionuclides onto Chitosan-acrylic acid-1-vinyl-2-vinylpyrrolidone/oxidized multi-walled carbon nanotubes (CTS-AA-VP/o-MWCNTs) composite has been investigated by batch adsorption technique. CTS-AA-VP/o-MWCNTs composite has been synthesized by copolymerization of acrylic acid (AA) and 1-vinyl-2-vinylpyrrolidone (VP) onto the surface of chitosan/oxidized multi-walled carbon nanotubes (CTS/o-MWCNTs) using gamma radiation. SEM, TGA, and FTIR were applied to characterize the morphology, thermal stability, and structure of the composite. The composite shows high removal capacity of 321.77, 369.91, and 456.46 mg/g towards 152+154Eu, 60Co, and 134Cs radionuclides, respectively.

Predicting Antibody Developability Profiles Through Early Stage Discovery Screening.

Monoclonal antibodies play an increasingly important role for the development of new drugs across multiple therapy areas. The term 'developability' encompasses the feasibility of molecules to successfully progress from discovery to development via evaluation of their physicochemical properties. These properties include the tendency for self-interaction and aggregation, thermal stability, colloidal stability, and optimization of their properties through sequence engineering. Selection of the best antibody molecule based on biological function, efficacy, safety, and developability allows for a streamlined and successful CMC phase. An efficient and practical high-throughput developability workflow (100 s-1,000 s of molecules) implemented during early antibody generation and screening is crucial to select the best lead candidates. This involves careful assessment of critical developability parameters, combined with binding affinity and biological properties evaluation using small amounts of purified material (<1 mg), as well as an efficient data management and database system. Herein, a panel of 152 various human or humanized monoclonal antibodies was analyzed in biophysical property assays. Correlations between assays for different sets of properties were established. We demonstrated in two case studies that physicochemical properties and key assay endpoints correlate with key downstream process parameters. The workflow allows the elimination of antibodies with suboptimal properties and a rank ordering of molecules for further evaluation early in the candidate selection process. This enables any further engineering for problematic sequence attributes without affecting program timelines.

Chitosan­g­maleic acid for effective removal of copper and nickel ions from their solutions.

Pollution of the environment associated with discharging the toxic heavy metals in water makes us focus the light to solve this problem. In continuation of our efforts, we aim in this work to utilize the graft copolymer (chitosan­g­maleic acid) to purify water from copper and nickel ions. The graft copolymer has been synthesized using gamma radiation and the grafting conditions have been optimized by studying the influence of acetic acid concentration (0.5-10% V/V), monomer concentration (5-17.5% w/v), chitosan concentration (0.25-2.5% w/v) and absorbed dose (0.5-5 kGy). FT-IR and TGA have been employed to characterize the graft copolymer. The metal ions uptake by the prepared graft copolymer was investigated and the influence of contact time, solution pH, polymer concentration, and metal ion concentration was studied. Adsorption kinetic models (pseudo-first-order, pseudo-second-order, and intra-particle diffusion equations) and adsorption isotherms (Langmuir, Freundlich, and Temkin equations) were also studied. It was found that the adsorption kinetics and isotherm agreed well with pseudo-second-order and Langmuir equations, respectively, indicating that the adsorption was chemisorption. The adsorption capacities of CTS­g­MA were 312.4 mg g-1 and 70.1 mg g-1 for Cu2+ and Ni2+, respectively. Effect of co-existence of other cationic ions on the adsorption capacity was also investigated.

Gamma radiation induced preparation of polyampholyte nanocomposite polymers for removal of Co(II).

A series of polyampholyte nanocomposite biopolymers, poly(N,N-diallyldimethylammonium chloride-co-acrylamide) grafted on carboxymethylcellulose/iron(III) oxide [P(DADMAC-AAm)CMC/Fe2O3] and poly(N,N-diallyldimethylammonium chloride-co-sodium acrylate) grafted on carboxymethylcellulose/iron(III) oxide [P(DADMAC-SA)CMC/Fe2O3], was prepared with different molar ratios of anionic groups to cationic groups using gamma irradiation. The grafting properties and swelling behavior were investigated as a function of grafting conditions such as DADMAC, AAm, SA, and CMC concentrations and absorbed dose. Fourier transform infrared spectroscopic analysis (FTIR) confirmed the graft copolymerization. Scanning electron microscope (SEM) was employed to check the morphological structure of CMC, P(DADMAC-AAm)CMC/Fe2O3, and P(DADMAC-SA)CMC/Fe2O3. Thermogravimetric analysis (TGA) and differential thermal analysis (DTA) further characterized the grafted copolymers and showed their high thermal stability. Using batch sorption experiments and 60Co as a radiotracer, P(DADMAC-AAm)CMC/Fe2O3 and P(DADMAC-SA)CMC/Fe2O3 were evaluated for Co(II) removal from aqueous solutions. Experimentally, P(DADMAC-AAm)CMC/Fe2O3 and P(DADMAC-SA)CMC/Fe2O3 show high sorption capacity of Co(II), i.e. 69.67 mg g-1 and 75.17 mg g-1, respectively, which makes them potential sorbents for Co(II) removal from water/wastewater. Finally, the Co(II) sorption was examined using sorption isotherm and kinetic models. Cobalt sorption was best fitted to Langmuir model which suggests the sorption is of chemisorption type. On the other hand, the sorption kinetics was best represented by the pseudo-first-order kinetic model.

Probing for submandibular gland peptide-T receptors on leukocytes with biotinylated-Lys-[Gly](6)-SGP-T.

Submandibular gland peptide-T (SGP-T) is a potent anti-chemotactic agent for human neutrophils possessing anti-inflammatory properties. Biologically active analogues of SGP-T have been synthesized and a biotinylated form (KG(6)-SGP-T; Bio-KG(6)-SGP-T) was utilized to identify binding sites on isolated human neutrophils. Neutrophils incubated with Bio-KG(6)-SGP-T followed by phycoerythrin (PE)-avidin secondary reagent were fixed and visualized using histochemistry and flow cytometry. At doses of 10(-8) and 10(-9) M, Bio-KG(6)-SGP-T was shown to bind to neutrophils. The binding of Bio-KG(6)-SGP-T, at doses of 10(-8) and 10(-9) M, to neutrophils was abolished by a 100-fold excess of non-biotinylated peptide (KG(6)-SGP-T), but not by 100-fold excess of SGP-T. However, all peptides, dose-dependently reduced the binding of a CD16b antibody (LNK16 clone) to isolated human neutrophils. This discrepancy probably results from different preferred conformations for Bio-KG(6)-SGP-T, KG(6)-SGP-T and SGP-T, since exhaustive conformational searches revealed a high degree of overlap between alpha-Bio-KG(6)-SGP-T and KG(6)-SGP-T that was not seen with SGP-T.

Identification of a binding site for the anti-inflammatory tripeptide feG.

The mechanism of action of feG, an anti-inflammatory peptide, was explored using data mining, molecular modeling, and enzymatic techniques. The molecular coordinates of protein kinase A (PKA) were used to create six virtual isoforms of protein kinase C (PKCalpha, betaI, betaII, delta, iota, and zeta). With in silico techniques a binding site for feG was identified on PKCbetaI that correlated significantly with a biological activity, the inhibition of intestinal anaphylaxis. Since feG selectively increased the binding of a PKCbetaI antibody, it is proposed that this peptide inhibits the reassociation of the hydrophobic tail of PKCbetaI with its binding site and prevents the enzyme from assuming an inactive conformation.

Submandibular gland tripeptide FEG (Phe-Glu-Gly) and analogues: keys to structure determination.

This study examined the structure activity relationship of NH(3)-Phe-Glu-Gly-COO(-) (FEG), a potent inhibitor of intestinal anaphylaxis. The inhibition by FEG analogues of antigen-provoked contractions of isolated ileal segments obtained from ovalbumin-sensitized rats was determined and molecular modeling performed. A combination of aromaticity of the first residue, minimal extension of the carboxyl group on residue 2, and underivatized N and C termini were essential for biological activity. FEG, WEG, WDG and the d-enantiomeric forms of FEG (feG) and YEG (yeG) retained biological activity. By considering dipole moments, the structural and conformational features critical to biological activity were established as the glutamyl-carboxyl group/Phe side chain and carboxyl/amino termini interactions. Analysis of Ramachandran plots for position 1 sidechains indicate that mobility of the aromatic sidechain must be restricted to retain biological activity. The anti-anaphylactic effects of FEG, characterized by specific structural and conformational restrictions, indicate a selective interaction with a receptor for this peptide in the intestine.

A tree-based algorithm for determining the effects of solvation on the structure of salivary gland tripeptide NH3+-D-PHE-D-GLU-GLY-COO-.

A D-enantiomeric analog of the submandibular gland rat-1 tripeptide FEG (Seq: NH(3)(+)-Phe-Glu-Gly-COO(-)) called feG (Seq: NH(3)(+)-D-Phe-D-Glu-Gly-COO(-)) was examined by molecular dynamics simulations in water. Previous in vacuo simulations suggested a conformation consisting predominantly of interactions between the Phe side chain and glutamyl-carboxyl group and a carboxyl/amino termini interaction. The solvated peptide was simulated using two approaches which were compared-a single 400-ns simulation and a "simulation tree." The "tree" approach utilized 45 10-ns simulations with different conformations used as initial structures for given trajectories. We demonstrate that multiple short duration simulations are able to describe the same conformational space as that described by longer simulations. Furthermore, previously described in vacuo interactions were confirmed with amendments: the previously described head-to-tail arrangement of the amino and carboxyl termini, was not observed; the interaction between the glutamyl carboxyl and Phe side chain describes only one of a continuum of conformations present wherein the aromatic residue remains in close proximity to the glutamyl carbonyl group, and also interacts with either of the two available carboxyl groups. Finally, utilizing only two separate 10-ns trajectories, we were able to better describe the conformational space than a single 60-ns trajectory, realizing a threefold decrease in the computational complexity of the problem.

Fast rotational matching of rigid bodies by fast Fourier transform acceleration of five degrees of freedom.

The 'fast rotational matching' method (an approach to find the three rotational degrees of freedom in matching problems using just one three-dimensional FFT) is extended to the full six-dimensional (rotation and translation) matching scenario between two three-dimensional objects. By recasting this problem into a formulation involving five angles and just one translational parameter, it was possible to accelerate, by means of fast Fourier transforms, five of the six degrees of freedom of the problem. This method was successfully applied to the docking of atomic structures of components into three-dimensional low-resolution density maps. Timing comparisons performed with our method and with 'fast translational matching' (the standard way to accelerate the translational parameters utilizing fast Fourier transforms) demonstrates that the performance gain can reach several orders of magnitude, especially for large map sizes. This gain can be particularly advantageous for spherical- and toroidal-shaped maps, since the scanning range of the translational parameter would be significantly constrained in these cases. The method can also be harnessed to the complementary surface (or 'exterior docking') problem and to pattern recognition in image processing.