RESUMO
We report here the creation of a graphical user interface (GUI) for the Data Extraction for Integrated Multidimensional Spectrometry (DEIMoS) tool. DEIMoS is a Python package that processes data from high-dimensional mass spectrometry measurements. It is divided into several modules, each representing a data processing step such as peak detection, alignment, and tandem mass spectra extraction and deconvolution. The inputs for and outputs from DEIMoS can include millions of N-dimensional data points, which can be challenging to visualize in a way that is interactive, informative, and responsive. Here, we used the HoloViz Python data visualization stack, including DataShader and Param, to create an interactive visualization of the mass spectrometry data. We believe the GUI will increase the accessibility of DEIMoS and that the visualization methods could be useful for other open-source mass spectrometry tools.
Assuntos
Software , Espectrometria de Massas/métodosRESUMO
A primary goal of metabolomics studies is to fully characterize the small-molecule composition of complex biological and environmental samples. However, despite advances in analytical technologies over the past two decades, the majority of small molecules in complex samples are not readily identifiable due to the immense structural and chemical diversity present within the metabolome. Current gold-standard identification methods rely on reference libraries built using authentic chemical materials ("standards"), which are not available for most molecules. Computational quantum chemistry methods, which can be used to calculate chemical properties that are then measured by analytical platforms, offer an alternative route for building reference libraries, i.e., in silico libraries for "standards-free" identification. In this review, we cover the major roadblocks currently facing metabolomics and discuss applications where quantum chemistry calculations offer a solution. Several successful examples for nuclear magnetic resonance spectroscopy, ion mobility spectrometry, infrared spectroscopy, and mass spectrometry methods are reviewed. Finally, we consider current best practices, sources of error, and provide an outlook for quantum chemistry calculations in metabolomics studies. We expect this review will inspire researchers in the field of small-molecule identification to accelerate adoption of in silico methods for generation of reference libraries and to add quantum chemistry calculations as another tool at their disposal to characterize complex samples.
Assuntos
Metabolômica , Teoria QuânticaRESUMO
We present DEIMoS: Data Extraction for Integrated Multidimensional Spectrometry, a Python application programming interface (API) and command-line tool for high-dimensional mass spectrometry data analysis workflows that offers ease of development and access to efficient algorithmic implementations. Functionality includes feature detection, feature alignment, collision cross section (CCS) calibration, isotope detection, and MS/MS spectral deconvolution, with the output comprising detected features aligned across study samples and characterized by mass, CCS, tandem mass spectra, and isotopic signature. Notably, DEIMoS operates on N-dimensional data, largely agnostic to acquisition instrumentation; algorithm implementations simultaneously utilize all dimensions to (i) offer greater separation between features, thus improving detection sensitivity, (ii) increase alignment/feature matching confidence among data sets, and (iii) mitigate convolution artifacts in tandem mass spectra. We demonstrate DEIMoS with LC-IMS-MS/MS metabolomics data to illustrate the advantages of a multidimensional approach in each data processing step.
Assuntos
Metabolômica , Espectrometria de Massas em Tandem , Algoritmos , Cromatografia Líquida/métodos , Metabolômica/métodos , Software , Espectrometria de Massas em Tandem/métodosRESUMO
The prediction of structure dependent molecular properties, such as collision cross sections as measured using ion mobility spectrometry, are crucially dependent on the selection of the correct population of molecular conformers. Here, we report an in-depth evaluation of multiple conformation selection techniques, including simple averaging, Boltzmann weighting, lowest energy selection, low energy threshold reductions, and similarity reduction. Generating 50â¯000 conformers each for 18 molecules, we used the In Silico Chemical Library Engine (ISiCLE) to calculate the collision cross sections for the entire data set. First, we employed Monte Carlo simulations to understand the variability between conformer structures as generated using simulated annealing. Then we employed Monte Carlo simulations to the aforementioned conformer selection techniques applied on the simulated molecular property: the ion mobility collision cross section. Based on our analyses, we found Boltzmann weighting to be a good trade-off between precision and theoretical accuracy. Combining multiple techniques revealed that energy thresholds and root-mean-squared deviation-based similarity reductions can save considerable computational expense while maintaining property prediction accuracy. Molecular dynamic conformer generation tools like AMBER can continue to generate new lowest energy conformers even after tens of thousands of generations, decreasing precision between runs. This reduced precision can be ameliorated and theoretical accuracy increased by running density functional theory geometry optimization on carefully selected conformers.
Assuntos
Espectrometria de Mobilidade Iônica , Simulação de Dinâmica Molecular , Conformação MolecularRESUMO
A growing number of software tools have been developed for metabolomics data processing and analysis. Many new tools are contributed by metabolomics practitioners who have limited prior experience with software development, and the tools are subsequently implemented by users with expertise that ranges from basic point-and-click data analysis to advanced coding. This Perspective is intended to introduce metabolomics software users and developers to important considerations that determine the overall impact of a publicly available tool within the scientific community. The recommendations reflect the collective experience of an NIH-sponsored Metabolomics Consortium working group that was formed with the goal of researching guidelines and best practices for metabolomics tool development. The recommendations are aimed at metabolomics researchers with little formal background in programming and are organized into three stages: (i) preparation, (ii) tool development, and (iii) distribution and maintenance.
Assuntos
Computação em Nuvem , Metabolômica/métodos , SoftwareRESUMO
The α2a adrenoceptor is a medically relevant subtype of the G protein-coupled receptor family. Unfortunately, high-throughput techniques aimed at producing novel drug leads for this receptor have been largely unsuccessful because of the complex pharmacology of adrenergic receptors. As such, cutting-edge in silico ligand- and structure-based assessment and de novo deep learning methods are well positioned to provide new insights into protein-ligand interactions and potential active compounds. In this work, we (i) collect a dataset of α2a adrenoceptor agonists and provide it as a resource for the drug design community; (ii) use the dataset as a basis to generate candidate-active structures via deep learning; and (iii) apply computational ligand- and structure-based analysis techniques to gain new insights into α2a adrenoceptor agonists and assess the quality of the computer-generated compounds. We further describe how such assessment techniques can be applied to putative chemical probes with a case study involving proposed medetomidine-based probes.
Assuntos
Aprendizado Profundo , Receptores Adrenérgicos alfa 2 , Ligantes , MedetomidinaRESUMO
We describe the Mass Spectrometry Adduct Calculator (MSAC), an automated Python tool to calculate the adduct ion masses of a parent molecule. Here, adduct refers to a version of a parent molecule [M] that is charged due to addition or loss of atoms and electrons resulting in a charged ion, for example, [M + H]+. MSAC includes a database of 147 potential adducts and adduct/neutral loss combinations and their mass-to-charge ratios (m/z) as extracted from the NIST/EPA/NIH Mass Spectral Library (NIST17), Global Natural Products Social Molecular Networking Public Spectral Libraries (GNPS), and MassBank of North America (MoNA). The calculator relies on user-selected subsets of the combined database to calculate expected m/z for adducts of molecules supplied as formulas. This tool is intended to help researchers create identification libraries to collect evidence for the presence of molecules in mass spectrometry data. While the included adduct database focuses on adducts typically detected during liquid chromatography-mass spectrometry analyses, users may supply their own lists of adducts and charge states for calculating expected m/z. We also analyzed statistics on adducts from spectra contained in the three selected mass spectral libraries. MSAC is freely available at https://github.com/pnnl/MSAC.
Assuntos
Espectrometria de Massas , Cromatografia Líquida/métodosRESUMO
Uncompetitive antagonists of the N-methyl d-aspartate receptor (NMDAR) have demonstrated therapeutic benefit in the treatment of neurological diseases such as Parkinson's and Alzheimer's, but some also cause dissociative effects that have led to the synthesis of illicit drugs. The ability to generate NMDAR antagonists in silico is therefore desirable for both new medication development and preempting and identifying new designer drugs. Recently, generative deep learning models have been applied to de novo drug design as a means to expand the amount of chemical space that can be explored for potential drug-like compounds. In this study, we assess the application of a generative model to the NMDAR to achieve two primary objectives: (i) the creation and release of a comprehensive library of experimentally validated NMDAR phencyclidine (PCP) site antagonists to assist the drug discovery community and (ii) an analysis of both the advantages conferred by applying such generative artificial intelligence models to drug design and the current limitations of the approach. We apply, and provide source code for, a variety of ligand- and structure-based assessment techniques used in standard drug discovery analyses to the deep learning-generated compounds. We present twelve candidate antagonists that are not available in existing chemical databases to provide an example of what this type of workflow can achieve, though synthesis and experimental validation of these compounds are still required.
Assuntos
Aprendizado Profundo , Receptores de N-Metil-D-Aspartato/antagonistas & inibidores , Bibliotecas de Moléculas Pequenas/química , Animais , Sítios de Ligação , Desenho de Fármacos , Ligantes , Camundongos , Estrutura Molecular , Receptores de N-Metil-D-Aspartato/química , Xenopus laevisRESUMO
Comprehensive and unambiguous identification of small molecules in complex samples will revolutionize our understanding of the role of metabolites in biological systems. Existing and emerging technologies have enabled measurement of chemical properties of molecules in complex mixtures and, in concert, are sensitive enough to resolve even stereoisomers. Despite these experimental advances, small molecule identification is inhibited by (i) chemical reference libraries (e.g., mass spectra, collision cross section, and other measurable property libraries) representing <1% of known molecules, limiting the number of possible identifications, and (ii) the lack of a method to generate candidate matches directly from experimental features (i.e., without a library). To this end, we developed a variational autoencoder (VAE) to learn a continuous numerical, or latent, representation of molecular structure to expand reference libraries for small molecule identification. We extended the VAE to include a chemical property decoder, trained as a multitask network, in order to shape the latent representation such that it assembles according to desired chemical properties. The approach is unique in its application to metabolomics and small molecule identification, with its focus on properties that can be obtained from experimental measurements (m/z, CCS) paired with its training paradigm, which involved a cascade of transfer learning iterations. First, molecular representation is learned from a large data set of structures with m/z labels. Next, in silico property values are used to continue training, as experimental property data is limited. Finally, the network is further refined by being trained with the experimental data. This allows the network to learn as much as possible at each stage, enabling success with progressively smaller data sets without overfitting. Once trained, the network can be used to predict chemical properties directly from structure, as well as generate candidate structures with desired chemical properties. Our approach is orders of magnitude faster than first-principles simulation for CCS property prediction. Additionally, the ability to generate novel molecules along manifolds, defined by chemical property analogues, positions DarkChem as highly useful in a number of application areas, including metabolomics and small molecule identification, drug discovery and design, chemical forensics, and beyond.
Assuntos
Simulação por Computador , Aprendizado Profundo , Bibliotecas de Moléculas Pequenas/análise , Metabolômica , Estrutura Molecular , Bibliotecas de Moléculas Pequenas/metabolismoRESUMO
High-throughput, comprehensive, and confident identifications of metabolites and other chemicals in biological and environmental samples will revolutionize our understanding of the role these chemically diverse molecules play in biological systems. Despite recent technological advances, metabolomics studies still result in the detection of a disproportionate number of features that cannot be confidently assigned to a chemical structure. This inadequacy is driven by the single most significant limitation in metabolomics, the reliance on reference libraries constructed by analysis of authentic reference materials with limited commercial availability. To this end, we have developed the in silico chemical library engine (ISiCLE), a high-performance computing-friendly cheminformatics workflow for generating libraries of chemical properties. In the instantiation described here, we predict probable three-dimensional molecular conformers (i.e., conformational isomers) using chemical identifiers as input, from which collision cross sections (CCS) are derived. The approach employs first-principles simulation, distinguished by the use of molecular dynamics, quantum chemistry, and ion mobility calculations, to generate structures and chemical property libraries, all without training data. Importantly, optimization of ISiCLE included a refactoring of the popular MOBCAL code for trajectory-based mobility calculations, improving its computational efficiency by over 2 orders of magnitude. Calculated CCS values were validated against 1983 experimentally measured CCS values and compared to previously reported CCS calculation approaches. Average calculated CCS error for the validation set is 3.2% using standard parameters, outperforming other density functional theory (DFT)-based methods and machine learning methods (e.g., MetCCS). An online database is introduced for sharing both calculated and experimental CCS values ( metabolomics.pnnl.gov ), initially including a CCS library with over 1 million entries. Finally, three successful applications of molecule characterization using calculated CCS are described, including providing evidence for the presence of an environmental degradation product, the separation of molecular isomers, and an initial characterization of complex blinded mixtures of exposure chemicals. This work represents a method to address the limitations of small molecule identification and offers an alternative to generating chemical identification libraries experimentally by analyzing authentic reference materials. All code is available at github.com/pnnl .
Assuntos
Quimioinformática/métodos , Teoria da Densidade Funcional , Bibliotecas de Moléculas Pequenas/química , Aprendizado de Máquina , Modelos Químicos , Simulação de Dinâmica MolecularRESUMO
The current gold standard for unambiguous molecular identification in metabolomics analysis is comparing two or more orthogonal properties from the analysis of authentic reference materials (standards) to experimental data acquired in the same laboratory with the same analytical methods. This represents a significant limitation for comprehensive chemical identification of small molecules in complex samples. The process is time consuming and costly, and the majority of molecules are not yet represented by standards. Thus, there is a need to assemble evidence for the presence of small molecules in complex samples through the use of libraries containing calculated chemical properties. To address this need, we developed a Multi-Attribute Matching Engine (MAME) and a library derived in part from our in silico chemical library engine (ISiCLE). Here, we describe an initial evaluation of these methods in a blinded analysis of synthetic chemical mixtures as part of the U.S. Environmental Protection Agency's (EPA) Non-Targeted Analysis Collaborative Trial (ENTACT, Phase 1). For molecules in all mixtures, the initial blinded false negative rate (FNR), false discovery rate (FDR), and accuracy were 57%, 77%, and 91%, respectively. For high evidence scores, the FDR was 35%. After unblinding of the sample compositions, we optimized the scoring parameters to better exploit the available evidence and increased the accuracy for molecules suspected as present. The final FNR, FDR, and accuracy were 67%, 53%, and 96%, respectively. For high evidence scores, the FDR was 10%. This study demonstrates that multiattribute matching methods in conjunction with in silico libraries may one day enable reduced reliance on experimentally derived libraries for building evidence for the presence of molecules in complex samples.
Assuntos
Biologia Computacional/métodos , Simulação por Computador , Bibliotecas de Moléculas Pequenas/química , Algoritmos , Bibliotecas de Moléculas Pequenas/metabolismoRESUMO
A series of Wrightia hanleyi extracts was screened for activity against Mycobacterium tuberculosis H37Rv. One active fraction contained a compound that initially appeared to be either the isoflavonoid wrightiadione or the alkaloid tryptanthrin, both of which have been previously reported in other Wrightia species. Characterization by NMR and MS, as well as evaluation of the literature describing these compounds, led to the conclusion that wrightiadione (1) was misidentified in the first report of its isolation from W. tomentosa in 1992 and again in 2015 when reported in W. pubescens and W. religiosa. Instead, the molecule described in these reports and in the present work is almost certainly the isobaric (same nominal mass) and isosteric (same number of atoms, valency, and shape) tryptanthrin (2), a well-known quinazolinone alkaloid found in a variety of plants including Wrightia species. Tryptanthrin (2) is also accessible synthetically via several routes and has been thoroughly characterized. Wrightiadione (1) has been synthesized and characterized and may have useful biological activity; however, this compound can no longer be said to be known to exist in Nature. To our knowledge, this misidentification of wrightiadione (1) has heretofore been unrecognized.
Assuntos
Antituberculosos/isolamento & purificação , Apocynaceae/química , Quinazolinas/isolamento & purificação , Antituberculosos/química , Antituberculosos/farmacologia , Espectroscopia de Ressonância Magnética Nuclear de Carbono-13 , Isoflavonas , Espectrometria de Massas , Testes de Sensibilidade Microbiana , Estrutura Molecular , Mycobacterium tuberculosis/efeitos dos fármacos , Espectroscopia de Prótons por Ressonância Magnética , Quinazolinas/química , Quinazolinas/farmacologiaRESUMO
BACKGROUND: Relatively small changes to gene expression data dramatically affect co-expression networks inferred from that data which, in turn, can significantly alter the subsequent biological interpretation. This error propagation is an underappreciated problem that, while hinted at in the literature, has not yet been thoroughly explored. Resampling methods (e.g. bootstrap aggregation, random subspace method) are hypothesized to alleviate variability in network inference methods by minimizing outlier effects and distilling persistent associations in the data. But the efficacy of the approach assumes the generalization from statistical theory holds true in biological network inference applications. RESULTS: We evaluated the effect of bootstrap aggregation on inferred networks using commonly applied network inference methods in terms of stability, or resilience to perturbations in the underlying expression data, a metric for accuracy, and functional enrichment of edge interactions. CONCLUSION: Bootstrap aggregation results in improved stability and, depending on the size of the input dataset, a marginal improvement to accuracy assessed by each method's ability to link genes in the same functional pathway.
Assuntos
Expressão Gênica/genética , Redes Reguladoras de Genes/genética , Algoritmos , HumanosRESUMO
CONTEXT: Computational fluid dynamics (CFD) simulations of airflows coupled with physiologically based pharmacokinetic (PBPK) modeling of respiratory tissue doses of airborne materials have traditionally used either steady-state inhalation or a sinusoidal approximation of the breathing cycle for airflow simulations despite their differences from normal breathing patterns. OBJECTIVE: Evaluate the impact of realistic breathing patterns, including sniffing, on predicted nasal tissue concentrations of a reactive vapor that targets the nose in rats as a case study. MATERIALS AND METHODS: Whole-body plethysmography measurements from a free-breathing rat were used to produce profiles of normal breathing, sniffing and combinations of both as flow inputs to CFD/PBPK simulations of acetaldehyde exposure. RESULTS: For the normal measured ventilation profile, modest reductions in time- and tissue depth-dependent areas under the curve (AUC) acetaldehyde concentrations were predicted in the wet squamous, respiratory and transitional epithelium along the main airflow path, while corresponding increases were predicted in the olfactory epithelium, especially the most distal regions of the ethmoid turbinates, versus the idealized profile. The higher amplitude/frequency sniffing profile produced greater AUC increases over the idealized profile in the olfactory epithelium, especially in the posterior region. CONCLUSIONS: The differences in tissue AUCs at known lesion-forming regions for acetaldehyde between normal and idealized profiles were minimal, suggesting that sinusoidal profiles may be used for this chemical and exposure concentration. However, depending upon the chemical, exposure system and concentration and the time spent sniffing, the use of realistic breathing profiles, including sniffing, could become an important modulator for local tissue dose predictions.
Assuntos
Modelos Biológicos , Respiração , Fenômenos Fisiológicos Respiratórios , Sistema Respiratório/metabolismo , Acetaldeído/farmacocinética , Animais , Feminino , Hidrodinâmica , Pletismografia Total , Ratos Sprague-DawleyRESUMO
Motivation: MerCat2 ("Mer-Catenate2") is a versatile, parallel, scalable and modular property software package for robustly analyzing features in omics data. Using massively parallel sequencing raw reads, assembled contigs, and protein sequences from any platform as input, MerCat2 performs k-mer counting of any length k, resulting in feature abundance counts tables, quality control reports, protein feature metrics, and graphical representation (i.e. principal component analysis (PCA)). Results: MerCat2 allows for direct analysis of data properties in a database-independent manner that initializes all data, which other profilers and assembly-based methods cannot perform. MerCat2 represents an integrated tool to illuminate omics data within a sample for rapid cross-examination and comparisons. Availability and implementation: MerCat2 is written in Python and distributed under a BSD-3 license. The source code of MerCat2 is freely available at https://github.com/raw-lab/mercat2. MerCat2 is compatible with Python 3 on Mac OS X and Linux. MerCat2 can also be easily installed using bioconda: mamba create -n mercat2 -c conda-forge -c bioconda mercat2.
RESUMO
Polycyclic aromatic hydrocarbons (PAHs) are a class of organic compounds frequently detected in the environment with widely varying toxicities. Many PAHs activate the aryl hydrocarbon receptor (AHR), inducing the expression of a battery of genes, including xenobiotic metabolizing enzymes like Cytochrome P450s (CYPs); however, not all PAHs act via this mechanism. We screened several parent and substituted PAHs in in vitro AHR activation assays to classify their unique activity. Retene (1-methyl-7-isopropylphenanthrene) displays Ahr2 dependent teratogenicity in zebrafish, but did not activate human AHR or zebrafish Ahr2, suggesting a retene metabolite activates Ahr2 in zebrafish to induce developmental toxicity. To investigate the role of metabolism in retene toxicity, studies were performed to determine the functional role of cyp1a, cyp1b1, and the microbiome in retene toxicity, identify the zebrafish window of susceptibility, and measure retene uptake, loss, and metabolite formation in vivo. Cyp1a-null fish were generated using CRISPR-Cas9. Cyp1a-null fish showed increased sensitivity to retene toxicity, while Cyp1b1-null fish were less susceptible, and microbiome elimination had no significant effect. Zebrafish required exposure to retene between 24 and 48 hours post fertilization (hpf) to exhibit toxicity. After static exposure, retene concentrations in zebrafish embryos increased until 24 hpf, peaked between 24 and 36 hpf, and decreased rapidly thereafter. We detected retene metabolites at 36 and 48 hpf, indicating metabolic onset preceding toxicity. This study highlights the value of combining molecular and systems biology approaches with mechanistic and predictive toxicology to interrogate the role of biotransformation in AHR-dependent toxicity.
RESUMO
Methods for assessing compound identification confidence in metabolomics and related studies have been debated and actively researched for the past two decades. The earliest effort in 2007 focused primarily on mass spectrometry and nuclear magnetic resonance spectroscopy and resulted in four recommended levels of metabolite identification confidence - the Metabolite Standards Initiative (MSI) Levels. In 2014, the original MSI Levels were expanded to five levels (including two sublevels) to facilitate communication of compound identification confidence in high resolution mass spectrometry studies. Further refinement in identification levels have occurred, for example to accommodate use of ion mobility spectrometry in metabolomics workflows, and alternate approaches to communicate compound identification confidence also have been developed based on identification points schema. However, neither qualitative levels of identification confidence nor quantitative scoring systems address the degree of ambiguity in compound identifications in context of the chemical space being considered, are easily automated, or are transferable between analytical platforms. In this perspective, we propose that the metabolomics and related communities consider identification probability as an approach for automated and transferable assessment of compound identification and ambiguity in metabolomics and related studies. Identification probability is defined simply as 1/N, where N is the number of compounds in a reference library or chemical space that match to an experimentally measured molecule within user-defined measurement precision(s), for example mass measurement or retention time accuracy, etc. We demonstrate the utility of identification probability in an in silico analysis of multi-property reference libraries constructed from the Human Metabolome Database and computational property predictions, provide guidance to the community in transparent implementation of the concept, and invite the community to further evaluate this concept in parallel with their current preferred methods for assessing metabolite identification confidence.
RESUMO
The use of anatomically accurate, animal-specific airway geometries is important for understanding and modeling the physiology of the respiratory system. One approach for acquiring detailed airway architecture is to create a bronchial cast of the conducting airways. However, typical casting procedures either do not faithfully preserve the in vivo branching angles or produce rigid casts that when removed for imaging are fragile and thus easily damaged. We address these problems by creating an in situ bronchial cast of the conducting airways in rats that can be subsequently imaged in situ using three-dimensional micro-CT imaging. We also demonstrate that deformations in airway branch angles resulting from the casting procedure are small, and that these angle deformations can be reversed through an interactive adjustment of the segmented cast geometry. Animal work was approved by the Institutional Animal Care and Use Committee of Pacific Northwest National Laboratory.
Assuntos
Brônquios/anatomia & histologia , Molde por Corrosão/métodos , Imageamento Tridimensional/métodos , Microtomografia por Raio-X/métodos , Animais , Artefatos , Broncografia , Masculino , Ratos , Ratos Sprague-DawleyRESUMO
The extrathoracic oral airway is not only a major mechanical barrier for pharmaceutical aerosols to reach the lung but also a major source of variability in lung deposition. Using computational fluid dynamics, deposition of 1−30 µm particles was predicted in 11 CT-based models of the oral airways of adults. Simulations were performed for mouth breathing during both inspiration and expiration at two steady-state flow rates representative of resting/nebulizer use (18 L/min) and of dry powder inhaler (DPI) use (45 L/min). Consistent with previous in vitro studies, there was a large intersubject variability in oral deposition. For an optimal size distribution of 1−5 µm for pharmaceutical aerosols, our data suggest that >75% of the inhaled aerosol is delivered to the intrathoracic lungs in most subjects when using a nebulizer but only in about half the subjects when using a DPI. There was no significant difference in oral deposition efficiency between inspiration and expiration, unlike subregional deposition, which shows significantly different patterns between the two breathing phases. These results highlight the need for incorporating a morphological variation of the upper airway in predictive models of aerosol deposition for accurate predictions of particle dosimetry in the intrathoracic region of the lung.
RESUMO
Computational methods for creating in silico libraries of molecular descriptors (e.g., collision cross sections) are becoming increasingly prevalent due to the limited number of authentic reference materials available for traditional library building. These so-called "reference-free metabolomics" methods require sampling sets of molecular conformers in order to produce high accuracy property predictions. Due to the computational cost of the subsequent calculations for each conformer, there is a need to sample the most relevant subset and avoid repeating calculations on conformers that are nearly identical. The goal of this study is to introduce a heuristic method of finding the most dissimilar conformers from a larger population in order to help speed up reference-free calculation methods and maintain a high property prediction accuracy. Finding the set of the n items most dissimilar from each other out of a larger population becomes increasingly difficult and computationally expensive as either n or the population size grows large. Because there exists a pairwise relationship between each item and all other items in the population, finding the set of the n most dissimilar items is different than simply sorting an array of numbers. For instance, if you have a set of the most dissimilar n = 4 items, one or more of the items from n = 4 might not be in the set n = 5. An exact solution would have to search all possible combinations of size n in the population exhaustively. We present an open-source software called similarity downselection (SDS), written in Python and freely available on GitHub. SDS implements a heuristic algorithm for quickly finding the approximate set(s) of the n most dissimilar items. We benchmark SDS against a Monte Carlo method, which attempts to find the exact solution through repeated random sampling. We show that for SDS to find the set of n most dissimilar conformers, our method is not only orders of magnitude faster, but it is also more accurate than running Monte Carlo for 1,000,000 iterations, each searching for set sizes n = 3-7 out of a population of 50,000. We also benchmark SDS against the exact solution for example small populations, showing that SDS produces a solution close to the exact solution in these instances. Using theoretical approaches, we also demonstrate the constraints of the greedy algorithm and its efficacy as a ratio to the exact solution.