RESUMEN
The polarisable machine-learned force field FFLUX requires pre-trained anisotropic Gaussian process regression (GPR) models of atomic energies and multipole moments to propagate unbiased molecular dynamics simulations. The outcome of FFLUX simulations is highly dependent on the predictive accuracy of the underlying models whose training entails determining the optimal set of model hyperparameters. Unfortunately, traditional direct learning (DL) procedures do not scale well on this task, especially when the hyperparameter search is initiated from a (set of) random guess solution(s). Additionally, the complexity of the hyperparameter space (HS) increases with the number of geometrical input features, at least for anisotropic kernels, making the optimization of hyperparameters even more challenging. In this study, we propose a transfer learning (TL) protocol that accelerates the training process of anisotropic GPR models by facilitating access to promising regions of the HS. The protocol is based on a seeding-relaxation mechanism in which an excellent guess solution is identified by rapidly building one or several small source models over a subset of the target training set before readjusting the previous guess over the entire set. We demonstrate the performance of this protocol by building and assessing the performance of DL and TL models of atomic energies and charges in various conformations of benzene, ethanol, formic acid dimer and the drug fomepizole. Our experiments suggest that TL models can be built one order of magnitude faster while preserving the quality of their DL analogs. Most importantly, when deployed in FFLUX simulations, TL models compete with or even outperform their DL analogs when it comes to performing FFLUX geometry optimization and computing harmonic vibrational modes.
RESUMEN
FFLUX is a quantum chemical topology-based multipolar force field that uses Gaussian process regression machine learning models to predict atomic energies and multipole moments on the fly for fast and accurate molecular dynamics simulations. These models have previously been trained on monomers, meaning that many-body effects, for example, intermolecular charge transfer, are missed in simulations. Moreover, dispersion and repulsion have been modeled using Lennard-Jones potentials, necessitating careful parametrization. In this work, we take an important step toward addressing these shortcomings and show that models trained on clusters, in this case, a dimer, can be used in FFLUX simulations by preparing and benchmarking a formamide dimer model. To mitigate the computational costs associated with training higher-dimensional models, we rely on the transfer of hyperparameters from a smaller source model to a larger target model, enabling an order of magnitude faster training than with a direct learning approach. The dimer model allows for simulations that account for two-body effects, including intermolecular polarization and charge penetration, and that do not require nonbonded potentials. We show that addressing these limitations allows for simulations that are closer to quantum mechanics than previously possible with the monomeric models.
RESUMEN
Predicting transition state geometries is one of the most challenging tasks in computational chemistry, which often requires expert-based knowledge and permanent human intervention. This short communication reports technical details and preliminary results of a python-based tool (AMADAR) designed to generate any Diels-Alder (DA) transition state geometry (TS) and analyze determined IRC paths in a (quasi-)automated fashion, given the product SMILES. Two modules of the package are devoted to performing, from IRC paths, reaction force analyses (RFA) and atomic (fragment) decompositions of the reaction force F and reaction force constant [Formula: see text]. The performance of the protocol has been assessed using a dataset of 2000 DA cycloadducts retrieved from the ZINC database. The sequential location of the corresponding TSs was achieved with a success rate of 95%. RFA plots confirmed the reaction force constant [Formula: see text] to be a good indicator of the (non)synchronicity of the associated DA reactions. Moreover, the atomic decomposition of [Formula: see text] allows for the rationalization of the (a)synchronicity of each DA reaction in terms of contributions stemming from pairs of interacting atoms. The source code of the AMADAR tool is available on GitHub [ CMCDD/AMADAR(github.com) ] and can be used directly with minor customizations, mostly regarding the local working environment of the user.
RESUMEN
Collection of DNA samples from subjects participating in clinical trials is vital to understanding variability in drug response. The purpose of this study was to assess pharmacogenetic sample-collection practices in the industry and to gather information on issues affecting collection. A survey questionnaire was developed and distributed to 20 pharmaceutical companies; 15 provided responses. Assessments included rate of DNA sample collection, reasons for low collection rates, reasons for rejection by health authorities (HAs) and institutional review boards/ethics committees (IRBs/ECs), and country-specific hurdles to sample collection. The results indicated that, although DNA samples are frequently collected, sample-acquisition rates remain lower than expected. Overall, the companies' experience has been that restrictions on sample usage are not consistently applied by regulatory bodies. This may reflect changing opinions/interpretations of HAs/IRBs/ECs. Collection of DNA samples in industry trials is still a challenge. Harmonization of sample-collection practices may facilitate the process.
Asunto(s)
Ensayos Clínicos como Asunto/métodos , ADN/análisis , Industria Farmacéutica/estadística & datos numéricos , Farmacogenética/métodos , Recolección de Datos , Humanos , Manejo de Especímenes/métodosRESUMEN
BACKGROUND: Parkinson disease (PD) is a common neurodegenerative disease affecting up to 1 million individuals in the United States. Depression affects up to 50% of these patients and is associated with a variety of poor outcomes for patients and their families. Despite this, there are few evidence-based data to guide clinical care. METHODS: An NIH-funded, randomized, controlled trial of paroxetine CR, nortriptyline, and placebo in 52 patients with PD and depression. The primary outcomes were the change in the Hamilton Depression Rating Scale (HAM-D) and the percentage of depression responders at 8 weeks. RESULTS: Nortriptyline was superior to placebo for the change in HAM-D (p < 0.002); paroxetine CR was not. There was a trend for superiority of nortriptyline over paroxetine CR at 8 weeks (p < 0.079). Response rates favored nortriptyline (p = 0.024): nortriptyline 53%, paroxetine CR 11%, placebo 24%. In planned contrasts of response rates, nortriptyline was superior to paroxetine CR (p = 0.034). Nortriptyline was also superior to placebo in many of the secondary outcomes, including sleep, anxiety, and social functioning, while paroxetine CR was not. Both active drug treatments were well tolerated. CONCLUSIONS: Though relatively modest in size, this is the largest placebo-controlled trial done to date in patients with Parkinson disease (PD) and depression. Nortriptyline was efficacious in the treatment of depression and paroxetine CR was not. When compared directly, nortriptyline produced significantly more responders than did paroxetine CR. The trial suggests that depression in patients with PD is responsive to treatment and raises questions about the relative efficacy of dual reuptake inhibitors and selective serotonin reuptake inhibitors.