Evaluating the chaos game representation of proteins for applications in machine learning models: prediction of antibody affinity and specificity as a case study.

CONTEXT: Machine learning techniques are becoming increasingly important in the selection and optimization of therapeutic molecules, as well as for the selection of formulation components and the prediction of long-term stability. Compared to first-principle models, machine learning techniques are easier to implement, and can identify correlations that would be hard to describe at a mechanistic level, but strongly rely on high-quality input training data. Here, we evaluate the potential of the "chaos game" representation to provide input data for machine learning models. The chaos game is an algorithm originally developed for the production of fractal structures, and later on applied also to the representation of biological sequences, such as genes and proteins. Our results show that the combination of the chaos game representation with convolutional neural networks results in comparable accuracy to other machine learning approaches, thus indicating that chaos game representations could be a valid alternative to existing featurization strategies for machine learning models of biological sequences. METHODS: We implement the chaos game in Python 3.8.10, and use it to produce fractal as well as novel expanding representations of protein sequences. We then feed the resulting images to a convolutional neural network, built in Python 3.8.10, using TensorFlow 2.9.1, Keras 2.9.0, and the scikit-learn 1.1.1 packages. We select as case study a recently published dataset for the antibody emibetuzumab, with the objective of co-optimizing antibodies variants with both high affinity and low non-specific binding.

Thermodynamic Modeling and Experimental Data Reveal That Sugars Stabilize Proteins According to an Excluded Volume Mechanism.

We present a new thermodynamic model to investigate the relative effects of excluded volume and soft interaction contributions in determining whether a cosolute will either destabilize or stabilize a protein in solution. This model is unique in considering an atomistically detailed model of the protein and accounting for the preferential accumulation/exclusion of the osmolyte molecules from the protein surface. Importantly, we use molecular dynamics simulations and experiments to validate the model. The experimental approach presents a unique means of decoupling excluded volume and soft interaction contributions using a linear polymeric series of cosolutes with different numbers of glucose subunits, from 1 (glucose) to 8 (maltooctaose), as well as an 8-mer of glucose units in the closed form (γ-CD). By studying the stabilizing effect of cosolutes along this polymeric series using lysozyme as a model protein, we validate the thermodynamic model and show that sugars stabilize proteins according to an excluded volume mechanism.

Mechanistic Investigation of tert-Butanol's Impact on Biopharmaceutical Formulations: When Experiments Meet Molecular Dynamics.

The use of tert-butyl alcohol for the lyophilization of pharmaceuticals has seen an uptick over the past years. Its advantages include increased solubility of hydrophobic drugs, enhanced product stability, shorter reconstitution time, and decreased processing time. While the mechanisms of protein stabilization exerted by cryo- and lyo-protectants are well known when water is the solvent of choice, little is known for organic solvents. This work investigates the interactions between two model proteins, namely, lactate dehydrogenase and myoglobin, and various excipients (mannitol, sucrose, 2-hydroxypropyl-ß-cyclodextrin and Tween 80) in the presence of tert-butyl alcohol. We thermally characterized mixtures of these components by differential scanning calorimetry and freeze-drying microscopy. We also spectroscopically evaluated the protein recovery after freezing and freeze-drying. We additionally performed molecular dynamics simulations to elucidate the interactions in ternary mixtures of the herein-investigated excipients, tert-butyl alcohol and the proteins. Both experiments and simulations revealed that tert-butyl alcohol had a detrimental impact on the recovery of the two investigated proteins, and no combination of excipients yielded a satisfactory recovery when the organic solvent was present within the formulation. Simulations suggested that the denaturing effect of tert-butyl alcohol was related to its propensity to accumulate in the proximity of the peptide surface, especially near positively charged residues.

Effect of Cosolutes on the Aggregation of a Tau Fragment: A Combined Experimental and Simulation Approach.

The intrinsically disordered protein Tau represents the main component of neurofibrillary tangles that are a hallmark of Alzheimer's disease. A small fragment of Tau, known as paired helical filament 6 (PHF6), is considered to be important for the formation of the ß-structure core of the fibrils. Here we study the aggregation of this fragment in the presence of different cosolutes, including urea, TMAO, sucrose and 2-hydroxypropyl-ß-cyclodextrin (2-HPßCD), using both experiments and molecular dynamics simulations. A novel implicit solvation approach (MIST - Model with Implicit Solvation Thermodynamics) is used, where an energetic contribution based on the concept of transfer free energies describes the effect of the cosolutes. The simulation predictions are compared to thioflavin-T and atomic force microscopy results, and the good agreement observed confirms the predictive ability of the computational approach herein proposed. Both simulations and experiments indicate that PHF6 aggregation is inhibited in the presence of urea and 2-HPßCD, while TMAO and sucrose stabilize associated conformations. The remarkable ability of HPßCD to inhibit aggregation represents an extremely promising result for future applications, especially considering the widespread use of this molecule as a drug carrier to the brain and as a solubilizer/excipient in pharmaceutical formulations.

Effect of diffusion kinetics on the ice nucleation temperature distribution.

The nucleation behavior of water is crucial in many fields, spanning meteorology, glaciology, biology, and astrophysics. We report observations suggesting an effect of diffusion kinetics in water on the heterogeneous immersion/contact mode nucleation temperature distribution of ice. We performed differential scanning calorimetry analyses of repeated freeze/thaw cycles and investigated the effect of several variables on the regularity of the nucleation temperature distributions obtained. We observed that the thawing temperature and residence time above 0 °C affect the width of the measured distributions. We explain the observed phenomena according to the diffusion behavior of an external nucleator. Specifically, conditions of enhanced diffusion of the nucleator translated into broader, more scattered distributions, while conditions of limited diffusion translated into narrower, more regular distributions. Lastly, based on our experimental findings, we propose a theoretical explanation centered on the temperature dependence of diffusion kinetics in water.

A New Transfer Free Energy Based Implicit Solvation Model for the Description of Disordered and Folded Proteins.

Most biological events occur on time scales that are difficult to access using conventional all-atom molecular dynamics simulations in explicit solvent. Implicit solvent techniques offer a promising solution to this problem, alleviating the computational cost associated with the simulation of large systems and accelerating the sampling compared to explicit solvent models. The substitution of water molecules by a mean field, however, introduces simplifications that may penalize accuracy and impede the prediction of certain physical properties. We demonstrate that existing implicit solvent models developed using a transfer free energy approach, while satisfactory at reproducing the folding behavior of globular proteins, fare less well in characterizing the conformational properties of intrinsically disordered proteins. We develop a new implicit solvent model that maximizes the degree of accuracy for both disordered and folded proteins. We show, by comparing the simulation outputs to experimental data, that in combination with the a99SB-disp force field, the implicit solvent model can describe both disordered (aß40, PaaA2, and drkN SH3) and folded ((AAQAA)3, CLN025, Trp-cage, and GTT) peptides. Our implicit solvent model permits a computationally efficient investigation of proteins containing both ordered and disordered regions, as well as the study of the transition between ordered and disordered protein states.

A Transfer Free Energy Based Implicit Solvent Model for Protein Simulations in Solvent Mixtures: Urea-Induced Denaturation as a Case Study.

We developed a method for implicit solvent molecular dynamics simulations of proteins in solvent mixtures (model with implicit solvation thermodynamics, MIST). The MIST method introduces experimental group transfer free energies to the generalized Born formulation for generating molecular trajectories without the need for developing rigorous explicit-solvent force fields for multicomponent solutions. As a test case, we studied the urea-induced denaturation of the Trp-cage miniprotein in water. We demonstrate that our method allows efficient exploration of the conformational space of the protein in only a few hundreds of nanoseconds of all-atom unbiased simulations. Furthermore, selective implementation of the transfer free energies of specific peptide groups, backbone, and side chains enables us to decouple their specific energetic contributions to the conformational changes of the protein. The approach herein developed can readily be extended to the investigation of complex matrices as well as to the characterization of protein aggregation. The MIST method is implemented in Plumed (ver. 2.8) as a separate module called SASA.

A proximity-based in silico approach to identify redox-labile disulfide bonds: The example of FVIII.

Allosteric disulfide bonds permit highly responsive, transient 'switch-like' properties that are ideal for processes like coagulation and inflammation that require rapid and localised responses to damage or injury. Haemophilia A (HA) is a rare bleeding disorder managed with exogenous coagulation factor(F) VIII products. FVIII has eight disulfide bonds and is known to be redox labile, but it is not known how reduction/oxidation affects the structure-function relationship, or its immunogenicity-a serious complication for 30% severe HA patients. Understanding how redox-mediated changes influence FVIII can inform molecular engineering strategies aimed at improving activity and stability, and reducing immunogenicity. FVIII is a challenging molecule to work with owing to its poor expression and instability so, in a proof-of-concept study, we used molecular dynamics (MD) to identify which disulfide bonds were most likely to be reduced and how this would affect structure/function; results were then experimentally verified. MD identified Cys1899-Cys1903 disulfide as the most likely to undergo reduction based on energy and proximity criteria. Further MD suggested this reduction led to a more open conformation. Here we present our findings and highlight the value of MD approaches.

Surface Treatment of Glass Vials for Lyophilization: Implications for Vacuum-Induced Surface Freezing.

Freeze-drying is commonly used to increase the shelf-life of pharmaceuticals and biopharmaceuticals. Freezing represents a crucial phase in the freeze-drying process, as it determines both cycle efficiency and product quality. For this reason, different strategies have been developed to allow for a better control of freezing, among them, the so-called vacuum-induced surface freezing (VISF), which makes it possible to trigger nucleation at the same time in all the vials being processed. We studied the effect of different vial types, characterized by the presence of hydrophilic (sulfate treatment) or hydrophobic (siliconization and TopLyo Si-O-C-H layer) inner coatings, on the application of VISF. We observed that hydrophobic coatings promoted boiling and blow-up phenomena, resulting in unacceptable aesthetic defects in the final product. In contrast, hydrophilic coatings increased the risk of fogging (i.e., the undesired creeping of the product upward along the inner vial surface). We also found that the addition of a surfactant (Tween 80) to the formulation suppressed boiling in hydrophobic-coated vials, but it enhanced the formation of bubbles. This undesired bubbling events induced by the surfactant could, however, be eliminated by a degassing step prior to the application of VISF. Overall, the combination of degasification and surfactant addition seems to be a promising strategy for the successful induction of nucleation by VISF in hydrophobic vials.

Force Field Parameterization for the Description of the Interactions between Hydroxypropyl-ß-Cyclodextrin and Proteins.

Cyclodextrins are cyclic oligosaccharides, widely used as drug carriers, solubilizers, and excipients. Among cyclodextrins, the functionalized derivative known as hydroxypropyl-ß-cyclodextrin (HPßCD) offers several advantages due to its unique structural features. Its optimal use in pharmaceutical and medical applications would benefit from a molecular-level understanding of its behavior, as can be offered by molecular dynamics simulations. Here, we propose a set of parameters for all-atom simulations of HPßCD, based on the ADD force field for sugars developed in our group, and compare it to the original CHARMM36 description. Using Kirkwood-Buff integrals of binary HPßCD-water mixtures as target experimental data, we show that the ADD-based description results in a considerably improved prediction of HPßCD self-association and interaction with water. We then use the new set of parameters to characterize the behavior of HPßCD toward the different amino acids. We observe pronounced interactions of HPßCD with both polar and nonpolar moieties, with a special preference for the aromatic rings of tyrosine, phenylalanine, and tryptophan. Interestingly, our simulations further highlight a preferential orientation of HPßCD's hydrophobic cavity toward the backbone atoms of amino acids, which, coupled with a favorable interaction of HPßCD with the peptide backbone, suggest a propensity for HPßCD to denature proteins.

Pressure Unfolding of Proteins: New Insights into the Role of Bound Water.

High pressures can be detrimental for protein stability, resulting in unfolding and loss of function. This phenomenon occurs because the unfolding transition is accompanied by a decrease in volume, which is typically attributed to the elimination of cavities that are present within the native state as a result of packing defects. We present a novel computational approach that enables the study of pressure unfolding in atomistically detailed protein models in implicit solvent. We include the effect of pressure using a transfer free energy term that allows us to decouple the effect of protein residues and bound water molecules on the volume change upon unfolding. We discuss molecular dynamics simulations results using this protocol for two model proteins, Trp-cage and staphylococcal nuclease (SNase). We find that the volume reduction of bound water is the key energetic term that drives protein denaturation under the effect of pressure, for both Trp-cage and SNase. However, we note differences in unfolding mechanisms between the smaller Trp-cage and the larger SNase protein. Indeed, the unfolding of SNase, but not Trp-cage, is seen to be further accompanied by a reduction in the volume of internal cavities. Our results indicate that, for small peptides, like Trp-cage, pressure denaturation is driven by the increase in solvent accessibility upon unfolding, and the subsequent increase in the number of bound water molecules. For larger proteins, like SNase, the cavities within the native fold act as weak spots, determining the overall resistance to pressure denaturation. Our simulations display a striking agreement with the pressure-unfolding profile experimentally obtained for SNase and represent a promising approach for a computationally efficient and accurate exploration of pressure-induced denaturation of proteins.

The Role of Cyclodextrins against Interface-Induced Denaturation in Pharmaceutical Formulations: A Molecular Dynamics Approach.

Protein-based pharmaceutical products are subject to a variety of environmental stressors, during both production and shelf-life. In order to preserve their structure, and, therefore, functionality, it is necessary to use excipients as stabilizing agents. Among the eligible stabilizers, cyclodextrins (CDs) have recently gained interest in the scientific community thanks to their properties. Here, a computational approach is proposed to clarify the role of ß-cyclodextrin (ßCD) and 2-hydroxypropyl-ß-cyclodextrin (HPßCD) against granulocyte colony-stimulating (GCSF) factor denaturation at the air-water and ice-water interfaces, and also in bulk water at 300 or 260 K. Both traditional molecular dynamics (MD) simulations and enhanced sampling techniques (metadynamics, MetaD) are used to shed light on the underlying molecular mechanisms. Bulk simulations revealed that CDs were preferentially included within the surface hydration layer of GCSF, and even included some peptide residues in their hydrophobic cavity. HPßCD was able to stabilize the protein against surface-induced denaturation in proximity of the air-water interface, while ßCD had a destabilizing effect. No remarkable conformational changes of GCSF, or noticeable effect of the CDs, were instead observed at the ice surface. GCSF seemed less stable at low temperature (260 K), which may be attributed to cold-denaturation effects. In this case, CDs did not significantly improve conformational stability. In general, the conformationally altered regions of GCSF seemed not to depend on the presence of excipients that only modulated the extent of destabilization with either a positive or a negative effect.

Protein Cold Denaturation in Implicit Solvent Simulations: A Transfer Free Energy Approach.

Proteins are stable over a narrow temperature range, with hot and cold denaturation occurring outside of this window, both of which adversely affect protein function. While hot unfolding is entropically driven, cold denaturation, on the other hand, results from a more favorable free energy associated with the interaction of water with apolar groups at low temperature. Because of the key role of water in this latter process, capturing cold denaturation using implicit solvent models is challenging. We propose here a novel computational approach to develop an implicit solvent model that accounts for both hot and cold denaturation in simulations involving atomistically detailed protein representations. By mining a large number of protein structures solved by nuclear magnetic resonance, we derive transfer free energy contributions for the backbone and amino acids side chains representing the transfer of these moieties between water at two different temperatures. Using Trp-cage as a model system, we show that the implicit solvent model constructed using these temperature-dependent free energies of transfer recovers the parabolic temperature dependence of protein stability, capturing both hot and cold denaturation. The resulting cold-unfolded conformations show reduced secondary structure content but preserve most of their internal hydrogen-bonding network, in contrast to the extended configurations with no hydrogen-bonding populated during heat-induced denaturation.

FVIII inhibitors display FV-neutralizing activity in the prothrombin time assay.

BACKGROUND: The coagulation factors (F)V and VIII are homologous proteins that support hemostasis through their regulation of FX activity. Hemophilia A (HA) patients have reduced FVIII activity and a prolonged bleeding time that is corrected through the administration of exogenous FVIII. Around one-third of severe HA patients develop FVIII neutralizing antibodies, known as "inhibitors," which neutralize FVIII activity and preclude them from further FVIII therapy. OBJECTIVES: We hypothesized that, based on the degree of homology between FV and FVIII (~40%), FVIII-neutralizing antibodies could cross react with FV. To test this hypothesis, a panel of recombinant, patient-derived, FVIII-neutralizing antibodies were screened for cross-reactivity against FV. METHODS: Factor V and FVIII activity was measured using one-stage clotting assays; structural analysis was carried out using a structural approach. RESULTS: We detected FV neutralizing activity with the anti-FVIII A2 domain antibody NB11B2. Because this antibody was derived from an HA inhibitor patient, FV-neutralizing activity was then evaluated in a number of HA inhibitor patient plasma samples; nine alloimmune samples had FV-neutralizing activity whereas no FV neutralizing activity was found in the two autoimmune samples available. We next examined the degree of surface homology between FV and FVIII and found that structural similarity could explain the cross reactivity of the anti-A2 antibody and likely accounts for the cross reactivity we observed in patient samples. CONCLUSIONS: Although this novel observation is of interest, further work will be needed to determine whether FV neutralization in HA patient samples contributes to their bleeding diathesis.

ADD Force Field for Sugars and Polyols: Predicting the Additivity of Protein-Osmolyte Interaction.

The protein-osmolyte interaction has been shown experimentally to follow an additive construct, where the individual osmolyte-backbone and osmolyte-side-chain interactions contribute to the overall conformational stability of proteins. Here, we computationally reconstruct this additive relation using molecular dynamics simulations, focusing on sugars and polyols, including sucrose and sorbitol, as model osmolytes. A new set of parameters (ADD) is developed for this purpose, using the individual Kirkwood-Buff integrals for sugar-backbone and sugar-side-chain interactions as target experimental data. We show that the ADD parameters can reproduce the additivity of protein-sugar interactions and correctly predict sucrose and sorbitol self-association and their interaction with water. The accurate description of the separate osmolyte-backbone and osmolyte-side-chain contributions also automatically translates into a good prediction of preferential exclusion from the surface of ribonuclease A and α-chymotrypsinogen A. The description of sugar polarity is improved compared to previous force fields, resulting in closer agreement with the experimental data and better compatibility with charged groups, such as the guanidinium moiety. The ADD parameters are developed in combination with the CHARMM36m force field for proteins, but good compatibility is also observed with the AMBER 99SB-ILDN and the OPLS-AA force fields. Overall, exploiting the additivity of protein-osmolyte interactions is a promising approach for the development of new force fields.

Terminal Capping of an Amyloidogenic Tau Fragment Modulates Its Fibrillation Propensity.

Aberrant protein folding leading to the formation of characteristic cross-ß-sheet-rich amyloid structures is well known for its association with a variety of debilitating human diseases. Often, depending upon amino acid composition, only a small segment of a large protein participates in amyloid formation and is in fact capable of self-assembling into amyloid, independent of the rest of the protein. Therefore, such peptide fragments serve as useful model systems for understanding the process of amyloid formation. An important factor that has often been overlooked while using peptides to mimic full-length protein is the charge on the termini of these peptides. Here, we show the influence of terminal charges on the aggregation of an amyloidogenic peptide from microtubule-associated protein Tau, implicated in Alzheimer's disease and tauopathies. We found that modification of terminal charges by capping the peptide at one or both of the termini drastically modulates the fibrillation of the hexapeptide sequence paired helical filament 6 (PHF6) from repeat 3 of Tau, both with and without heparin. Without heparin, the PHF6 peptide capped at both termini and PHF6 capped only at the N-terminus self-assembled to form amyloid fibrils. With heparin, all capping variants of PHF6, except for PHF6 with both termini free, formed typical amyloid fibrils. However, the rate and extent of aggregation both with and without heparin as well as the morphology of aggregates were found to be highly dependent on the terminal charges. Our molecular dynamics simulations on PHF6 capping variants corroborated our experiments and provided critical insights into the mechanism of PHF6 self-assembly. Overall, our results emphasize the importance of terminal modifications in fibrillation of small peptide fragments and provide significant insights into the aggregation of a small Tau fragment, which is considered essential for Tau filament assembly.

A model-based approach for the rational design of the freeze-thawing of a protein-based formulation.

Proteins are unstable molecules that may be severely injured by stresses encountered during freeze-thawing. Despite this, the selection of freeze-thaw conditions is currently empirical, and this results in reduced process control. Here we propose a mathematical model that takes into account the leading causes of protein instability during freeze-thawing, i.e. cold denaturation and surface-induced unfolding, and may guide the selection of optimal operating conditions. It is observed that a high cooling rate is beneficial for molecules that are extremely sensitive to cold denaturation, while the opposite is true when ice-induced unfolding is dominant. In all cases, a fast thawing rate is observed to be beneficial. The simulation outputs are confirmed by experimental data for myoglobin and lactate dehydrogenase, suggesting that the proposed modeling approach can reproduce the main features of protein behavior during freeze-thawing. This approach can therefore guide the selection of optimal conditions for protein-based formulations that are stored in a frozen or freeze-dried state.

Impact of controlled vacuum induced surface freezing on the freeze drying of human plasma.

During the freezing step of a typical freeze drying process, the temperature at which nucleation is induced is generally stochastically distributed, resulting in undesired within-batch heterogeneity. Controlled nucleation techniques have been developed to address this problem; these make it possible to trigger the formation of ice crystals at the same time and temperature in all the batch. Here, the controlled nucleation technique known as vacuum induced surface freezing is compared to spontaneous freezing for the freeze drying of human plasma, a highly concentrated system commonly stored in a dried state. The potency of Factor VIII (FVIII), a sensitive, labile protein present in plasma, and the reconstitution time of the dried cakes are evaluated immediately after freeze drying, and after 1, 3, 6 or 9 months storage at different degradation temperatures. We show that the application of controlled nucleation significantly reduces the reconstitution time and in addition helps to improve FVIII stability.

The Ice-Water Interface and Protein Stability: A Review.

The ice-water interface is commonly encountered in our life, and comes into play in a wide number of natural phenomena. Here, attention will be focused on its effects on protein stability, with specific reference to the case of pharmaceutical proteins. This field represents a fascinating, and not yet fully understood, subject of investigation. Some background information on the ice-water phase diagram, as well as to the mechanisms of nucleation and crystal growth, will be provided. We will eventually discuss the effect of ice on protein activity, reviewing the mechanisms of ice-induced denaturation that have been proposed so far and discussing the strategies that may help prevent, or minimize, undesired loss of therapeutic activity.

Heightened Cold-Denaturation of Proteins at the Ice-Water Interface.

The process of freezing proteins is widely used in applications ranging from processing and storage of biopharmaceuticals to cryo-EM analysis of protein complexes. The formation of an ice-water interface is a critical destabilization factor for the protein, which can be offset by the use of cryo-protectants. Using molecular dynamics simulation, we demonstrate that the presence of the ice-water interface leads to a lowering of the free-energy barrier for unfolding, resulting in rapid unfolding of the protein. The unfolding process does not require direct adsorption of the protein to the surface, but is rather mediated by nearby liquid molecules that show an increased tendency for hydrating nonpolar groups. The observed enhancement in the cold denaturation process upon ice formation can be mitigated by addition of glucose, which acts as a cryoprotectant through preferential exclusion from side chains of the protein.