Multi-Task Learning of Scanning Electron Microscopy and Synthetic Thermal Tomography Images for Detection of Defects in Additively Manufactured Metals.

One of the key challenges in laser powder bed fusion (LPBF) additive manufacturing of metals is the appearance of microscopic pores in 3D-printed metallic structures. Quality control in LPBF can be accomplished with non-destructive imaging of the actual 3D-printed structures. Thermal tomography (TT) is a promising non-contact, non-destructive imaging method, which allows for the visualization of subsurface defects in arbitrary-sized metallic structures. However, because imaging is based on heat diffusion, TT images suffer from blurring, which increases with depth. We have been investigating the enhancement of TT imaging capability using machine learning. In this work, we introduce a novel multi-task learning (MTL) approach, which simultaneously performs the classification of synthetic TT images, and segmentation of experimental scanning electron microscopy (SEM) images. Synthetic TT images are obtained from computer simulations of metallic structures with subsurface elliptical-shaped defects, while experimental SEM images are obtained from imaging of LPBF-printed stainless-steel coupons. MTL network is implemented as a shared U-net encoder between the classification and the segmentation tasks. Results of this study show that the MTL network performs better in both the classification of synthetic TT images and the segmentation of SEM images tasks, as compared to the conventional approach when the individual tasks are performed independently of each other.

Hotspot Identification and Drug Design of Protein-Protein Interaction Modulators Using the Fragment Molecular Orbital Method.

Protein-protein interactions (PPIs) are essential for the function of many proteins. Aberrant PPIs have the potential to lead to disease, making PPIs promising targets for drug discovery. There are over 64,000 PPIs in the human interactome reference database; however, to date, very few PPI modulators have been approved for clinical use. Further development of PPI-specific therapeutics is highly dependent on the availability of structural data and the existence of reliable computational tools to explore the interface between two interacting proteins. The fragment molecular orbital (FMO) quantum mechanics method offers comprehensive and computationally inexpensive means of identifying the strength (in kcal/mol) and the chemical nature (electrostatic or hydrophobic) of the molecular interactions taking place at the protein-protein interface. We have integrated FMO and PPI exploration (FMO-PPI) to identify the residues that are critical for protein-protein binding (hotspots). To validate this approach, we have applied FMO-PPI to a dataset of protein-protein complexes representing several different protein subfamilies and obtained FMO-PPI results that are in agreement with published mutagenesis data. We observed that critical PPIs can be divided into three major categories: interactions between residues of two proteins (intermolecular), interactions between residues within the same protein (intramolecular), and interactions between residues of two proteins that are mediated by water molecules (water bridges). We extended our findings by demonstrating how this information obtained by FMO-PPI can be utilized to support the structure-based drug design of PPI modulators (SBDD-PPI).

In-silico guided discovery of novel CCR9 antagonists.

Antagonism of CCR9 is a promising mechanism for treatment of inflammatory bowel disease, including ulcerative colitis and Crohn's disease. There is limited experimental data on CCR9 and its ligands, complicating efforts to identify new small molecule antagonists. We present here results of a successful virtual screening and rational hit-to-lead campaign that led to the discovery and initial optimization of novel CCR9 antagonists. This work uses a novel data fusion strategy to integrate the output of multiple computational tools, such as 2D similarity search, shape similarity, pharmacophore searching, and molecular docking, as well as the identification and incorporation of privileged chemokine fragments. The application of various ranking strategies, which combined consensus and parallel selection methods to achieve a balance of enrichment and novelty, resulted in 198 virtual screening hits in total, with an overall hit rate of 18%. Several hits were developed into early leads through targeted synthesis and purchase of analogs.

Predictions of Ligand Selectivity from Absolute Binding Free Energy Calculations.

Binding selectivity is a requirement for the development of a safe drug, and it is a critical property for chemical probes used in preclinical target validation. Engineering selectivity adds considerable complexity to the rational design of new drugs, as it involves the optimization of multiple binding affinities. Computationally, the prediction of binding selectivity is a challenge, and generally applicable methodologies are still not available to the computational and medicinal chemistry communities. Absolute binding free energy calculations based on alchemical pathways provide a rigorous framework for affinity predictions and could thus offer a general approach to the problem. We evaluated the performance of free energy calculations based on molecular dynamics for the prediction of selectivity by estimating the affinity profile of three bromodomain inhibitors across multiple bromodomain families, and by comparing the results to isothermal titration calorimetry data. Two case studies were considered. In the first one, the affinities of two similar ligands for seven bromodomains were calculated and returned excellent agreement with experiment (mean unsigned error of 0.81 kcal/mol and Pearson correlation of 0.75). In this test case, we also show how the preferred binding orientation of a ligand for different proteins can be estimated via free energy calculations. In the second case, the affinities of a broad-spectrum inhibitor for 22 bromodomains were calculated and returned a more modest accuracy (mean unsigned error of 1.76 kcal/mol and Pearson correlation of 0.48); however, the reparametrization of a sulfonamide moiety improved the agreement with experiment.

Rapid and accurate assessment of GPCR-ligand interactions Using the fragment molecular orbital-based density-functional tight-binding method.

The reliable and precise evaluation of receptor-ligand interactions and pair-interaction energy is an essential element of rational drug design. While quantum mechanical (QM) methods have been a promising means by which to achieve this, traditional QM is not applicable for large biological systems due to its high computational cost. Here, the fragment molecular orbital (FMO) method has been used to accelerate QM calculations, and by combining FMO with the density-functional tight-binding (DFTB) method we are able to decrease computational cost 1000 times, achieving results in seconds, instead of hours. We have applied FMO-DFTB to three different GPCR-ligand systems. Our results correlate well with site directed mutagenesis data and findings presented in the published literature, demonstrating that FMO-DFTB is a rapid and accurate means of GPCR-ligand interactions. © 2017 Authors. Journal of Computational Chemistry Published by Wiley Periodicals, Inc.

Using the fragment molecular orbital method to investigate agonist-orexin-2 receptor interactions.

The understanding of binding interactions between any protein and a small molecule plays a key role in the rationalization of affinity and selectivity and is essential for an efficient structure-based drug discovery (SBDD) process. Clearly, to begin SBDD, a structure is needed, and although there has been fantastic progress in solving G-protein-coupled receptor (GPCR) crystal structures, the process remains quite slow and is not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of X-ray crystallography to impact the drug discovery process for GPCR targets in 'real-time' and hence there is still a need for other practical and cost-efficient alternatives. We present here an approach that integrates our previously described hierarchical GPCR modelling protocol (HGMP) and the fragment molecular orbital (FMO) quantum mechanics (QM) method to explore the interactions and selectivity of the human orexin-2 receptor (OX2R) and its recently discovered nonpeptidic agonists. HGMP generates a 3D model of GPCR structures and its complexes with small molecules by applying a set of computational methods. FMO allowsab initioapproaches to be applied to systems that conventional QM methods would find challenging. The key advantage of FMO is that it can reveal information on the individual contribution and chemical nature of each residue and water molecule to the ligand binding that normally would be difficult to detect without QM. We illustrate how the combination of both techniques provides a practical and efficient approach that can be used to analyse the existing structure-function relationships (SAR) and to drive forward SBDD in a real-world example for which there is no crystal structure of the complex available.

The Fragment Molecular Orbital Method Reveals New Insight into the Chemical Nature of GPCR-Ligand Interactions.

Our interpretation of ligand-protein interactions is often informed by high-resolution structures, which represent the cornerstone of structure-based drug design. However, visual inspection and molecular mechanics approaches cannot explain the full complexity of molecular interactions. Quantum Mechanics approaches are often too computationally expensive, but one method, Fragment Molecular Orbital (FMO), offers an excellent compromise and has the potential to reveal key interactions that would otherwise be hard to detect. To illustrate this, we have applied the FMO method to 18 Class A GPCR-ligand crystal structures, representing different branches of the GPCR genome. Our work reveals key interactions that are often omitted from structure-based descriptions, including hydrophobic interactions, nonclassical hydrogen bonds, and the involvement of backbone atoms. This approach provides a more comprehensive picture of receptor-ligand interactions than is currently used and should prove useful for evaluation of the chemical nature of ligand binding and to support structure-based drug design.

Beyond Membrane Protein Structure: Drug Discovery, Dynamics and Difficulties.

Most of the previous content of this book has focused on obtaining the structures of membrane proteins. In this chapter we explore how those structures can be further used in two key ways. The first is their use in structure based drug design (SBDD) and the second is how they can be used to extend our understanding of their functional activity via the use of molecular dynamics. Both aspects now heavily rely on computations. This area is vast, and alas, too large to consider in depth in a single book chapter. Thus where appropriate we have referred the reader to recent reviews for deeper assessment of the field. We discuss progress via the use of examples from two main drug target areas; G-protein coupled receptors (GPCRs) and ion channels. We end with a discussion of some of the main challenges in the area.

Design, synthesis and structure-activity relationships of a novel class of sulfonylpyridine inhibitors of Interleukin-2 inducible T-cell kinase (ITK).

Starting from benzylpyrimidine 2, molecular modeling and X-ray crystallography were used to design highly potent inhibitors of Interleukin-2 inducible T-cell kinase (ITK). Sulfonylpyridine 4i showed sub-nanomolar affinity against ITK, was selective versus Lck and its activity in the Jurkat cell-based assay was greatly improved over 2.

Unsupervised learning-enabled pulsed infrared thermographic microscopy of subsurface defects in stainless steel.

Metallic structures produced with laser powder bed fusion (LPBF) additive manufacturing method (AM) frequently contain microscopic porosity defects, with typical approximate size distribution from one to 100 microns. Presence of such defects could lead to premature failure of the structure. In principle, structural integrity assessment of LPBF metals can be accomplished with nondestructive evaluation (NDE). Pulsed infrared thermography (PIT) is a non-contact, one-sided NDE method that allows for imaging of internal defects in arbitrary size and shape metallic structures using heat transfer. PIT imaging is performed using compact instrumentation consisting of a flash lamp for deposition of a heat pulse, and a fast frame infrared (IR) camera for measuring surface temperature transients. However, limitations of imaging resolution with PIT include blurring due to heat diffusion, sensitivity limit of the IR camera. We demonstrate enhancement of PIT imaging capability with unsupervised learning (UL), which enables PIT microscopy of subsurface defects in high strength corrosion resistant stainless steel 316 alloy. PIT images were processed with UL spatial-temporal separation-based clustering segmentation (STSCS) algorithm, refined by morphology image processing methods to enhance visibility of defects. The STSCS algorithm starts with wavelet decomposition to spatially de-noise thermograms, followed by UL principal component analysis (PCA), fine-tuning optimization, and neural learning-based independent component analysis (ICA) algorithms to temporally compress de-noised thermograms. The compressed thermograms were further processed with UL-based graph thresholding K-means clustering algorithm for defects segmentation. The STSCS algorithm also includes online learning feature for efficient re-training of the model with new data. For this study, metallic specimens with calibrated microscopic flat bottom hole defects, with diameters in the range from 203 to 76 µm, were produced using electro discharge machining (EDM) drilling. While the raw thermograms do not show any material defects, using STSCS algorithm to process PIT images reveals defects as small as 101 µm in diameter. To the best of our knowledge, this is the smallest reported size of a sub-surface defect in a metal imaged with PIT, which demonstrates the PIT capability of detecting defects in the size range relevant to quality control requirements of LPBF-printed high-strength metals.

Toward an understanding of agonist binding to human Orexin-1 and Orexin-2 receptors with G-protein-coupled receptor modeling and site-directed mutagenesis.

The class A G-protein-coupled receptors (GPCRs) Orexin-1 (OX1) and Orexin-2 (OX2) are located predominantly in the brain and are linked to a range of different physiological functions, including the control of feeding, energy metabolism, modulation of neuro-endocrine function, and regulation of the sleep-wake cycle. The natural agonists for OX1 and OX2 are two neuropeptides, Orexin-A and Orexin-B, which have activity at both receptors. Site-directed mutagenesis (SDM) has been reported on both the receptors and the peptides and has provided important insight into key features responsible for agonist activity. However, the structural interpretation of how these data are linked together is still lacking. In this work, we produced and used SDM data, homology modeling followed by MD simulation, and ensemble-flexible docking to generate binding poses of the Orexin peptides in the OX receptors to rationalize the SDM data. We also developed a protein pairwise similarity comparing method (ProS) and a GPCR-likeness assessment score (GLAS) to explore the structural data generated within a molecular dynamics simulation and to help distinguish between different GPCR substates. The results demonstrate how these newly developed methods of structural assessment for GPCRs can be used to provide a working model of neuropeptide-Orexin receptor interaction.

Structure-based design and synthesis of potent benzothiazole inhibitors of interleukin-2 inducible T cell kinase (ITK).

Inhibition of the non-receptor tyrosine kinase ITK, a component of the T-cell receptor signalling cascade, may represent a novel treatment for allergic asthma. Here we report the structure-based optimization of a series of benzothiazole amides that demonstrate sub-nanomolar inhibitory potency against ITK with good cellular activity and kinase selectivity. We also elucidate the binding mode of these inhibitors by solving the X-ray crystal structures of several inhibitor-ITK complexes.

Fighting obesity with a sugar-based library: discovery of novel MCH-1R antagonists by a new computational-VAST approach for exploration of GPCR binding sites.

Obesity is an increasingly common disease. While antagonism of the melanin-concentrating hormone-1 receptor (MCH-1R) has been widely reported as a promising therapeutic avenue for obesity treatment, no MCH-1R antagonists have reached the market. Discovery and optimization of new chemical matter targeting MCH-1R is hindered by reduced HTS success rates and a lack of structural information about the MCH-1R binding site. X-ray crystallography and NMR, the major experimental sources of structural information, are very slow processes for membrane proteins and are not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of these methods to impact the drug discovery process for GPCR targets in "real-time", and hence, there is an urgent need for other practical and cost-efficient alternatives. We present here a conceptually pioneering approach that integrates GPCR modeling with design, synthesis, and screening of a diverse library of sugar-based compounds from the VAST technology (versatile assembly on stable templates) to provide structural insights on the MCH-1R binding site. This approach creates a cost-efficient new avenue for structure-based drug discovery (SBDD) against GPCR targets. In our work, a primary VAST hit was used to construct a high-quality MCH-1R model. Following model validation, a structure-based virtual screen yielded a 14% hit rate and 10 novel chemotypes of potent MCH-1R antagonists, including EOAI3367472 (IC50 = 131 nM) and EOAI3367474 (IC50 = 213 nM).

Physics-informed neural network with transfer learning (TL-PINN) based on domain similarity measure for prediction of nuclear reactor transients.

Nuclear reactor safety and efficiency can be enhanced through the development of accurate and fast methods for prediction of reactor transient (RT) states. Physics informed neural networks (PINNs) leverage deep learning methods to provide an alternative approach to RT modeling. Applications of PINNs in monitoring of RTs for operator support requires near real-time model performance. However, as with all machine learning models, development of a PINN involves time-consuming model training. Here, we show that a transfer learning (TL-PINN) approach achieves significant performance gain, as measured by reduction of the number of iterations for model training. Using point kinetic equations (PKEs) model with six neutron precursor groups, constructed with experimental parameters of the Purdue University Reactor One (PUR-1) research reactor, we generated different RTs with experimentally relevant range of variables. The RTs were characterized using Hausdorff and Fréchet distance. We have demonstrated that pre-training TL-PINN on one RT results in up to two orders of magnitude acceleration in prediction of a different RT. The mean error for conventional PINN and TL-PINN models prediction of neutron densities is smaller than 1%. We have developed a correlation between TL-PINN performance acceleration and similarity measure of RTs, which can be used as a guide for application of TL-PINNs.

Study of human Orexin-1 and -2 G-protein-coupled receptors with novel and published antagonists by modeling, molecular dynamics simulations, and site-directed mutagenesis.

The class A G-protein-coupled receptors (GPCRs) Orexin-1 (OX1) and Orexin-2 (OX2) are located predominantly in the brain and are linked to a range of different physiological functions, including the control of feeding, energy metabolism, modulation of neuro-endocrine function, and regulation of the sleep-wake cycle. Site-directed mutagenesis (SDM) and domain exchange (chimera) studies have provided important insight into key features of the OX1 and OX2 binding sites. However, the precise determinants of antagonist binding and selectivity are still not fully known. In this work, we used homology modeling of OX receptors to direct further SDM studies. These SDM studies were followed by molecular dynamics (MD) simulations to rationalize the full scope of the SDM data and to explain the role of each mutated residue in the binding and selectivity of a set of OX antagonists: Almorexant (dual OX1 and OX2 antagonist), SB-674042 (OX1 selective antagonist), EMPA (OX2 selective antagonist), and others. Our primary interest was focused on transmembrane helix 3 (TM3), which is identified as being of great importance for the selectivity of OX antagonists. These studies revealed conformational differences between the TM3 helices of OX1 and OX2, resulting from differences in amino acid sequences of the OX receptors that affect key interhelical interactions formed between TM3 and neighboring TM domains. The MD simulation protocol used here, which was followed by flexible docking studies, went beyond the use of static models and allowed for a more detailed exploration of the OX structures. In this work, we have demonstrated how even small differences in the amino acid sequences of GPCRs can lead to significant differences in structure, antagonist binding affinity, and selectivity of these receptors. The MD simulations allowed refinement of the OX receptor models to a degree that was not possible with static homology modeling alone and provided a deeper rationalization of the SDM data obtained. To validate these findings and to demonstrate that they can be usefully applied to the design of novel, very selective OX antagonists, we show here two examples of antagonists designed in house: EP-109-0092 (OX1 selective) and EP-009-0513 (OX2 selective).

Predicting Residence Time of GPCR Ligands with Machine Learning.

Drug-target residence time, the duration of binding at a given protein target, has been shown in some protein families to be more significant for conferring efficacy than binding affinity. To carry out efficient optimization of residence time in drug discovery, machine learning models that can predict that value need to be developed. One of the main challenges with predicting residence time is the paucity of data. This chapter outlines all of the currently available ligand kinetic data, providing a repository that contains the largest publicly available source of GPCR-ligand kinetic data to date. To help decipher the features of kinetic data that might be beneficial to include in computational models for the prediction of residence time, the experimental evidence for properties that influence residence time are summarized. Finally, two different workflows for predicting residence time with machine learning are outlined. The first is a single-target model trained on ligand features; the second is a multi-target model trained on features generated from molecular dynamics simulations.

Fighting COVID-19 with Artificial Intelligence.

The development of vaccines for the treatment of COVID-19 is paving the way for new hope. Despite this, the risk of the virus mutating into a vaccine-resistant variant still persists. As a result, the demand of efficacious drugs to treat COVID-19 is still pertinent. To this end, scientists continue to identify and repurpose marketed drugs for this new disease. Many of these drugs are currently undergoing clinical trials and, so far, only one has been officially approved by FDA. Drug repurposing is a much faster route to the clinic than standard drug development of novel molecules, nevertheless in a pandemic this process is still not fast enough to halt the spread of the virus. Artificial intelligence has already played a large part in hastening the drug discovery process, not only by facilitating the selection of potential drug candidates but also in monitoring the pandemic and enabling faster diagnosis of patients. In this chapter, we focus on the impact and challenges that artificial intelligence has demonstrated thus far with respect to drug repurposing of therapeutics for the treatment of COVID-19.

Novel 6,7,8,9-tetrahydro-5H-1,4,7,10a-tetraaza-cyclohepta[f]indene analogues as potent and selective 5-HT(2C) agonists for the treatment of metabolic disorders.

The discovery of a novel series of 5-HT(2C) agonists based on a tricyclic pyrazolopyrimidine scaffold is described. Compounds with good levels of in vitro potency and moderate to good levels of selectivity with respect to the 5-HT(2A) and 5-HT(2B) receptors were identified. One of the analogues (7 g) was found to be efficacious in a sub-chronic weight loss model. A key limitation of the series of compounds was that they were found to be potent inhibitors of the hERG ion channel. Some compounds, bearing polar side chains were identified which showed a much reduced hERG liability however these compounds were sub-optimal in terms of their in vitro potency or selectivity.

Optical methodology for detecting histologically unapparent nanoscale consequences of genetic alterations in biological cells.

Recently, there has been a major thrust to understand biological processes at the nanoscale. Optical microscopy has been exceedingly useful in imaging cell microarchitecture. Characterization of cell organization at the nanoscale, however, has been stymied by the lack of practical means of cell analysis at these small scales. To address this need, we developed a microscopic spectroscopy technique, single-cell partial-wave spectroscopy (PWS), which provides insights into the statistical properties of the nanoscale architecture of biological cells beyond what conventional microscopy reveals. Coupled with the mesoscopic light transport theory, PWS quantifies the disorder strength of intracellular architecture. As an illustration of the potential of the technique, in the experiments with cell lines and an animal model of colon carcinogenesis we show that increase in the degree of disorder in cell nanoarchitecture parallels genetic events in the early stages of carcinogenesis in otherwise microscopically/histologically normal-appearing cells. These data indicate that this advance in single-cell optics represented by PWS may have significant biomedical applications.

Characterizing Rhodopsin-Arrestin Interactions with the Fragment Molecular Orbital (FMO) Method.

Arrestin binding to G protein-coupled receptors (GPCRs) plays a vital role in receptor signaling. Recently, the crystal structure of rhodopsin bound to activated visual arrestin was resolved using XFEL (X-ray free electron laser). However, even with the crystal structure in hand, our ability to understand GPCR-arrestin binding is limited by the availability of accurate tools to explore receptor-arrestin interactions. We applied fragment molecular orbital (FMO) method to explore the interactions formed between the residues of rhodopsin and arrestin. FMO enables ab initio approaches to be applied to systems that conventional quantum mechanical (QM) methods would be too compute-expensive. The FMO calculations detected 35 significant interactions involved in rhodopsin-arrestin binding formed by 25 residues of rhodopsin and 28 residues of arrestin. Two major regions of interaction were identified: at the C-terminal tail of rhodopsin (D330-S343) and where the "finger loop" (G69-T79) of arrestin directly inserts into rhodopsin active core. Out of these 35 interactions, 23 were mainly electrostatic and 12 hydrophobic in nature.