AutoSolvate: A Toolkit for Automating Quantum Chemistry Design and Discovery of Solvated Molecules

The availability of large, high-quality data sets is crucial for artificial intelligence design and discovery in chemistry. Despite the essential roles of solvents in chemistry, the rapid computational data set generation of solution-phase molecular properties at the quantum mechanical level of theory was previously hampered by the complicated simulation procedure. Software toolkits that can automate the procedure to set up high-throughput explicit-solvent quantum chemistry (QC) calculations for arbitrary solutes and solvents in an open-source framework are still lacking. We developed AutoSolvate, an open-source toolkit to streamline the workflow for QC calculation of explicitly solvated molecules. It automates the solvated-structure generation, force field fitting, configuration sampling, and the final extraction of microsolvated cluster structures that QC packages can readily use to predict molecular properties of interest. AutoSolvate is available through both a command line interface and a graphical user interface, making it accessible to the broader scientific community. To improve the quality of the initial structures generated by AutoSolvate, we investigated the dependence of solute-solvent closeness on solute/solvent identities and trained a machine learning model to predict the closeness and guide initial structure generation. Finally, we tested the capability of AutoSolvate for rapid data set curation by calculating the outer-sphere reorganization energy of a large data set of 166 redox couples, which demonstrated the promise of the AutoSolvate package for chemical discovery efforts.

TOPOLOGY OPTIMIZATION OF LIGHTWEIGHT STRUCTURES WITH APPLICATION TO BONE SCAFFOLDS AND 3D PRINTED SHOES FOR DIABETICS

An automatic complex topology lightweight structure generation method (ACTLSGM) is presented to automatically generate 3D models of lightweight truss structures with a boundary surface of any shape. The core idea of the ACTLSGM is to use the PIMesh, a mesh generation algorithm developed by the authors, to generate node distributions inside the object representing the boundary surface of the target complex topology structures; raw lightweight truss structures are then generated based on the node distributions; the resulting lightweight truss structure is then created by adjusting the radius of the raw truss structures using an optimization algorithm based on finite element truss analysis. The finite element analysis-based optimization algorithm can ensure the resulting structures satisfy the design requirements on stress distributions or stiffness. Three demos, including a lightweight structure for a cantilever beam, a femur bone scaffold, and a 3D shoe sole model with adaptive stiffness that can be used to adjust foot pressure distributions for patients with diabetic foot problems, are generated to demonstrate the performance of the ACTLSGM. The ACTLSGM is not limited to generating 3D models of medical devices, but can be applied in many other fields, including 3D printing infills and other fields where customized lightweight structures are required.

HyFactor: Hydrogen-count labelled graph-based defactorization Autoencoder

Graph-based architectures are becoming increasingly popular as a tool for structure generation. Here, we introduce a novel open-source architecture HyFactor which is inspired by previously reported DEFactor architecture and based on the hydrogen labeled graphs. Since the original DEFactor code was not available, its new implementation (ReFactor) was prepared in this work for the benchmarking purpose. HyFactor demonstrates its high performance on the ZINC 250K MOSES and ChEMBL data set and in molecular generation tasks, it is considerably more effective than ReFactor. The code of HyFactor and all models obtained in this study are publicly available from our GitHub repository: https://github.com/Laboratoire-de-Chemoinformatique/hyfactor

On the Online Coalition Structure Generation Problem

We consider the online version of the coalition structure generation problem, in which agents, corresponding to the vertices of a graph, appear in an online fashion and have to be partitioned into coalitions by an authority (i.e., an online algorithm). When an agent appears, the algorithm has to decide whether to put the agent into an existing coalition or to create a new one containing, at this moment, only her. The decision is irrevocable. The objective is partitioning agents into coalitions so as to maximize the resulting social welfare that is the sum of all coalition values. We consider two cases for the value of a coalition: (1) the sum of the weights of its edges, and (2) the sum of the weights of its edges divided by its size. Coalition structures appear in a variety of application in AI, multi-agent systems, networks, as well as in social networks, data analysis, computational biology, game theory, and scheduling. For each of the coalition value functions we consider the bounded and unbounded cases depending on whether or not the size of a coalition can exceed a given value α. Furthermore, we consider the case of a limited number of coalitions and various weight functions for the edges, i.e., unrestricted, positive and constant weights. We show tight or nearly tight bounds for the competitive ratio in each case.

Why Can't Individuals with Nonfluent Aphasia Produce Sentences? Exploring the Role of Lexical Availability

<p>Nonfluent aphasia is a language disorder characterised by sparse, fragmented speech. Individuals with this disorder often produce single words accurately (for example, they can name pictured objects), but have great difficulty producing sentences. An important research goal is to understand why sentences are so difficult for these individuals. To produce a sentence, a speaker must not only retrieve its lexical elements, but also integrate them into a grammatically well-formed sentence. Indeed, most research to date has focused on this grammatical integration process. However, recent studies suggest that the noun and/or verb content of the sentence can also be an important determinant of success (e.g., Raymer & Kohen, 2006; Speer & Wilshire, 2014). In this thesis, I explore the role of noun availability on sentence production accuracy using an identity priming paradigm. Participants are asked to describe a pictured event using a single sentence (e.g., “The fish is kissing the turtle”). In the critical condition, an auditory prime word is presented just prior to the picture, which is identical to one of the nouns in the target sentence (e.g., fish). The rationale is that the prime will enhance the availability of its counterpart when the person comes to produce the target sentence. Participants were four individuals with mild nonfluent aphasia, two individuals with fluent aphasia, and six older, healthy controls. Consistent with our hypotheses, the nonfluent participants as a group were more accurate at producing sentences when one of its nouns – either the subject or object - was primed in this way. Importantly, in the primed subject noun condition, these results held even when accuracy on the primed element itself was excluded, suggesting it had a broad effect on sentence production accuracy. The primed nouns had no effect on sentence production accuracy for the fluent individuals or the controls. We interpret these findings within models of sentence production that allow for considerable interplay between the processes of lexical content retrieval and sentence structure generation.</p>

Why Can't Individuals with Nonfluent Aphasia Produce Sentences? Exploring the Role of Lexical Availability

<p>Nonfluent aphasia is a language disorder characterised by sparse, fragmented speech. Individuals with this disorder often produce single words accurately (for example, they can name pictured objects), but have great difficulty producing sentences. An important research goal is to understand why sentences are so difficult for these individuals. To produce a sentence, a speaker must not only retrieve its lexical elements, but also integrate them into a grammatically well-formed sentence. Indeed, most research to date has focused on this grammatical integration process. However, recent studies suggest that the noun and/or verb content of the sentence can also be an important determinant of success (e.g., Raymer & Kohen, 2006; Speer & Wilshire, 2014). In this thesis, I explore the role of noun availability on sentence production accuracy using an identity priming paradigm. Participants are asked to describe a pictured event using a single sentence (e.g., “The fish is kissing the turtle”). In the critical condition, an auditory prime word is presented just prior to the picture, which is identical to one of the nouns in the target sentence (e.g., fish). The rationale is that the prime will enhance the availability of its counterpart when the person comes to produce the target sentence. Participants were four individuals with mild nonfluent aphasia, two individuals with fluent aphasia, and six older, healthy controls. Consistent with our hypotheses, the nonfluent participants as a group were more accurate at producing sentences when one of its nouns – either the subject or object - was primed in this way. Importantly, in the primed subject noun condition, these results held even when accuracy on the primed element itself was excluded, suggesting it had a broad effect on sentence production accuracy. The primed nouns had no effect on sentence production accuracy for the fluent individuals or the controls. We interpret these findings within models of sentence production that allow for considerable interplay between the processes of lexical content retrieval and sentence structure generation.</p>

Leveraging molecular structure and bioactivity with chemical language models for drug design

Generative chemical language models (CLMs) can be used for de novo molecular structure generation. These CLMs learn from the structural information of known molecules to generate new ones. In this paper, we show that “hybrid” CLMs can additionally leverage the bioactivity information available for the training compounds. To computationally design ligands of phosphoinositide 3-kinase gamma (PI3Kγ), we created a large collection of virtual molecules with a generative CLM. This primary virtual compound library was further refined using a CLM-based classifier for bioactivity prediction. This second hybrid CLM was pretrained with patented molecular structures and fine-tuned with known PI3Kγ binders and non-binders by transfer learning. Several of the computer-generated molecular designs were commercially available, which allowed for fast prescreening and preliminary experimental validation. A new PI3Kγ ligand with sub-micromolar activity was identified. The results positively advocate hybrid CLMs for virtual compound screening and activity-focused molecular design in low-data situations.

A High-Similarity Modeling Method for Low-Porosity Porous Material and Its Application in Bearing Cage Self-Lubrication Simulation

The porous oil-containing cage achieves the storage, spillage, and suction of lubricating oil by its micro-pore structure, thus ensuring the self-lubricating performance of the bearing. Carrying out fast and accurate modeling of the cage microscopic pore structure is the key to the analysis of the self-lubricating mechanism of bearings. In response to the issues where current modeling methods of porous materials have a low similarity of pore distribution, morphology, structure, and size characteristics, and the transition of pore surfaces is sharp, this paper proposed a modeling method of a highly similar micro-pore structure based on the idea of median filtering, the quartet structure generation set (QSGS), and the slice method. By extracting and analyzing the pore characteristics of the porous model and comparing them with the experimental results of CT scanning, the advantages of the modeling method in terms of morphology and pore connectivity were verified. Finally, by carrying out simulation analysis of the centrifugal force of oil splashing and capillary oil absorption on the constructed model by combining the parameters of porous structures such as porosity and tortuosity, the advantages of the modeling method in the construction of the porous model and multi-physical field analysis were further verified.

Stochastic-based liquid apparent permeability model of shale oil reservoir considering geological control

Shale is a complex porous medium composed of organic matter (OM) and inorganic minerals (iOM). Because of its widespread nanopores, using Darcy’s law is challenging. In this work, a two-fluid system model is established to calculate the oil flow rate in a single nanopore. Then, a spatial distribution model of shale components is constructed with a modified quartet structure generation set algorithm. The stochastic apparent permeability (AP) model of shale oil is finally established by combining the two models. The proposed model can consider the effects of various geological controls: the content and grain size distribution of shale components, pore size distribution, pore types and nanoconfined effects (slip length and spatially varying viscosity). The results show that slip length in OM nanopores is far greater than that in iOM. However, when the total organic content is less than 0.3 ~ 0.4, the effect of the OM slip on AP increases first and then decreases with the decrease in mean pore size, resulting in that the flow enhancement in shale is much smaller than that in a single nanopore. The porosity distribution and grain size distribution are also key factors affecting AP. If we ignore the difference of porosity between shale components, the error of permeability estimation is more than 200%. Similarly, the relative error can reach 20% if the effect of grain size distribution is ignored. Our model can help understand oil transport in shale strata and provide parameter characterization for numerical simulation.