Estimation Of Distribution Research Articles

Orbital Exenteration in Recurrence Cancer: 5 Years Experience.

The purpose of this study was to assess the overall survival (OS) and disease-free survival (DFS) of patients who underwent orbital exenteration for periorbital, conjunctival, and primary intraorbital carcinomas. Additionally, we assessed the outcomes of anterior retrograde temporalis muscle flap restoration. For all patients who had orbital exenteration in the previous five years, a non-comparative retrospective assessment of their medical records, histology, and radiographic imaging was carried out. We investigated the relationships between the various qualitative factors using Cramer's V Kaplan-Meier (KM) analysis. For each of the patient's categorical factors that were of relevance, estimates of the survival distribution were displayed, and log-rank tests were used to determine whether the survival distributions were equal. This study looks at 19 participants. The sample is made up of 13 men (68%) and 6 women (32%). The degree of relationship (Cramer's V index) between lymph node metastases (N) and the existence of distant metastases (M) is high, at 64%, and is statistically significant because the p-value is 0.0034 < 0.005. Lymph node metastases had a statistically significant impact on overall survival (p = 0.04 < 0.05). Thirteen of the nineteen patients tested had no palsy (68%). There was no one presenting a CSF leak. Our findings show how crucial it is to identify any lymph node involvement that orbital neoplasms may have. In patients who have received many treatments, sentinel lymph node biopsy (SLNB) may be used to determine the stage and spread of the cancer. To determine whether additional tumor characteristics may be explored, more expertise in the SLNB field for patients with orbital cancer who have received many treatments may be helpful. To prevent additional scarring and to be comparable to previous techniques for facial nerve lesions, the anterior retrograde approach and the transorbital procedure for temporal muscle flap in-setting are both effective methods.

Estimating absorption time distributions of general Markov jump processes

AbstractThe estimation of absorption time distributions of Markov jump processes is an important task in various branches of statistics and applied probability. While the time‐homogeneous case is classic, the time‐inhomogeneous case has recently received increased attention due to its added flexibility and advances in computational power. However, commuting sub‐intensity matrices are assumed, which in various cases limits the parsimonious properties of the resulting representation. This paper develops the theory required to solve the general case through maximum likelihood estimation, and in particular, using the expectation‐maximization algorithm. A reduction to a piecewise constant intensity matrix function is proposed in order to provide succinct representations, where a parametric linear model binds the intensities together. Practical aspects are discussed and illustrated through the estimation of notoriously demanding theoretical distributions and real data, from the perspective of matrix analytic methods.

Indoor radon interval prediction in the Swedish building stock using machine learning

Indoor radon represents a health hazard for occupants. However, the indoor radon measurement rate is low in Sweden because of no mandatory requirements. Measuring indoor radon on an urban scale is complicated, machine learning exploiting existing data for pattern identification provides a cost-efficient approach to estimate indoor radon exposure in the building stock. Extreme gradient boosting (XGBoost) models and deep neural network (DNN) models were developed based on indoor radon measurement records, property registers, and geogenic information. The XGBoost models showed promising results in predicting indoor radon intervals for different types of buildings with macro-F1 between 0.93 and 0.96, whereas the DNN models attained macro-F1 between 0.64 and 0.74. After that, the XGBoost models trained on the national indoor radon dataset were transferred to fit building registers in metropolitan regions to estimate the indoor radon intervals in non-measured and measured buildings by regions and building classes. By comparing the prediction results and the statistical summary of indoor radon intervals in measured buildings, the model uncertainty and validity were determined. The study ascertains the prediction performance of machine learning models in classifying indoor radon intervals and discusses the benefits and limitations of the data-driven approach. The research outcomes can assist preliminary large-scale indoor radon distribution estimation for relevant authorities and guide onsite measurements for prioritized building stock prone to indoor radon exposure.

Dynamic Source Distribution and Emission Inventory of a Persistent, Mobile, and Toxic (PMT) Substance, Melamine, in China.

Persistent, mobile, and toxic (PMT) substances are affecting the safety of drinking water and are threatening the environment and human health. Many PMT substances are used in industrial processing or consumer products, but their sources and emissions mostly remain unclear. This study presents a long-term source distribution and emission estimation of melamine, a high-production-volume PMT substance of emerging global concern. The results indicate that in China, approximately 1858.7 kilotonnes (kt) of melamine were released into the water (∼58.9%), air (∼27.0%), and soil systems (∼14.1%) between 1995 and 2020, mainly from its production and use in the decorative panels, textiles, and paper industries. The textile and paper industries have the highest emission-to-consumption ratios, with more than 90% emissions per unit consumption. Sewage treatment plants are the largest source of melamine in the environment for the time being, but in-use products and their wastes will serve as significant melamine sources in the future. The study prompts priority action to control the risk of PMT substances internationally.

Estimation of cumulative amplitude distributions of miniature postsynaptic currents allows characterising their multimodality, quantal size and variability

A miniature postsynaptic current (mPSC) is a small, rare, and highly variable spontaneous synaptic event that is generally caused by the spontaneous release of single vesicles. The amplitude and variability of mPSCs are key measures of the postsynaptic processes and are taken as the main characteristics of an elementary unit (quantal size) in traditional quantal analysis of synaptic transmission. Due to different sources of biological and measurement noise, recordings of mPSCs exhibit high trial-to-trial heterogeneity, and experimental measurements of mPSCs are usually noisy and scarce, making their analysis demanding. Here, we present a sequential procedure for precise analysis of mPSC amplitude distributions for the range of small currents. To illustrate the developed approach, we chose previously obtained experimental data on the effect of the extracellular matrix on synaptic plasticity. The proposed statistical technique allowed us to identify previously unnoticed additional modality in the mPSC amplitude distributions, indicating the formation of new immature synapses upon ECM attenuation. We show that our approach can reliably detect multimodality in the distributions of mPSC amplitude, allowing for accurate determination of the size and variability of the quantal synaptic response. Thus, the proposed method can significantly expand the informativeness of both existing and newly obtained experimental data. We also demonstrated that mPSC amplitudes around the threshold of microcurrent excitation follow the Gumbel distribution rather than the binomial statistics traditionally used for a wide range of currents, either for a single synapse or when taking into consideration small influences of the adjacent synapses. Such behaviour is argued to originate from the theory of extreme processes. Specifically, recorded mPSCs represent instant random current fluctuations, among which there are relatively larger spikes (extreme events). They required more level of coherence that can be provided by different mechanisms of network or system level activation including neuron circuit signalling and extrasynaptic processes.

First estimate of distribution, abundance, and risk of encounter with aquaculture vessels for the rare Chilean dolphin in the entire Northern Chilean Patagonia

Abstract Assessing distribution and abundance patterns for rare species is challenging yet imperative, considering the extant, potentially hazardous, anthropogenic activities. In particular, the poorly studied Chilean dolphin (Cephalorhynchus eutropia) exhibits an extremely patchy distribution and low densities, co‐occurring with the intensive aquaculture industry in Southern Chile, among other anthropogenic activities. The distribution and abundance patterns of Chilean dolphins were assessed in the entire Northern Chilean Patagonia. A hierarchical species distribution model was fitted to data from line transect surveys using distance sampling techniques and environmental variables derived from topographic features and oceanographic models. A second version of this model used a joint‐likelihood approach to incorporate presence–pseudoabsence data for improving parameter estimation. Spatial predictions arising from these models were used to evaluate the relative probability of a dolphin encountering vessels from the local, predominantly aquaculture fleet. The results show that using a joint‐likelihood approach drastically reduced uncertainty in the parameters controlling the effect of covariates and in the total abundance estimates. This model estimated an overall low abundance (median 2,225.8; 95% CI 1,340–3,867) for this species in the region, and indicates their preference for shallow, sheltered bays and inner channels, near river mouths, where low salinity is expected. The highest probabilities of dolphin–vessel interactions were found on the eastern coast of Chiloe Big Island, coinciding with the largest number of aquaculture concessions in the area. Considering that the overall population of the Chilean dolphin is expected to be in the low thousands, that suitable habitat for the species is highly restricted, and that the species is facing increasing anthropogenic impacts, with some areas undergoing or planning major expansions in development, the results provided here should be considered in management plans for extant marine protected areas, and in the evaluation of current International Union for Conservation of Nature (IUCN) and national conservation categories for the species.

Spatial distribution, source identification and flux estimation of polycyclic aromatic hydrocarbons and organochlorine pesticides in basins of the Eastern Indian Ocean

Although polycyclic aromatic hydrocarbons (PAHs) and organochlorine pesticides (OCPs) have been recorded worldwide, information on their presence in the Eastern Indian Ocean (EIO), especially south of 10°S, remains limited. We investigated the distribution and depositional fluxes of PAHs and OCPs, and the major sources and ecological risks of PAHs in EIO surface sediments from the Central Indian Ocean (CIOB) and Wharton Basin (WB). The concentration of Σ18 PAHs and ∑10 OCPs had an average value (± SD) of 138.4 ± 52.34 and 0.8 ± 0.20 ng g−1, respectively. PAHs may mainly affected by traffic emission and biomass and wood combustion. Persistent organic pollutant accumulation rate (PAR) and depositional flux (DF) values showed that abundant PAHs might lost during top-down transport. The low trans- chordane (CHL)/cis-CHL ratio and PAR of OCPs may indicated few OCPs were inputted into the EIO recently. The results of binary isotope mixing modeling indicate the predominance of marine organic matter (MOM) in total organic carbon (TOC) of sediments. Fluoranthene (Flour) and pyrene (Py) might have potential biological effects in the EIO. The study provided background values for PAHs and OCPs in the Indian Ocean, and preliminarily revealed the fate of POPs in the open oceans.

A distribution-free method for probabilistic prediction

Machine learning techniques have been widely used and are mostly performed by predicting point estimation. Nevertheless, there are various scenarios that require more information beyond only point estimation. Probabilistic prediction is a typical research topic that provides probability distribution while predicting, which addressed with much attention recently. Most of the previous works make an assumption on the probabilistic distribution to a certain extent, which may lead to potential errors in the subsequent decision-making procedure. In this paper, we propose a distribution-free method for regression problems on real-value response under the probabilistic prediction framework and present an effective boosting-based method to perform the training process. Moreover, we introduce a further improved method to accomplish an unbiased mean estimation of the target distribution. Thorough experiments on multiple benchmark data are conducted to demonstrate the effectiveness of the proposed method with regard to different measures.

A Back Propagation Neural Network Model for Postharvest Blueberry Shelf-Life Prediction Based on Feature Selection and Dung Beetle Optimizer

(1) Background: Traditional kinetic-based shelf-life prediction models have low fitting accuracy and inaccurate prediction results for blueberries. Therefore, this study aimed to develop a blueberry shelf-life prediction method based on a back propagation neural network (BPNN) optimized by the dung beetle optimizer using an elite pool strategy and a Gaussian distribution estimation strategy (GDEDBO); (2) Methods: The “Liberty” blueberry cultivar was used as the research object, and 23 quality indicators, including color parameters, weight loss rate, decay rate, and texture parameters, were measured under storage temperatures of 0, 4, and 25 °C. Based on the maximum relevance minimum redundancy (MRMR) algorithm, seven key influencing factors of shelf life were selected as the input parameters of the model, and then the MRMR-GDEDBO-BPNN prediction model was established; (3) Results: the results showed that the model outperformed the baseline model at all three temperatures, with strong generalization ability, high prediction accuracy, and reliability; and (4) Conclusions: this study provided a theoretical basis for the shelf-life determination of blueberries under different storage temperatures and offered technical support for the prediction of remaining shelf life.

Spatial patterns and environmental benefits of photovoltaic poverty alleviation programs in China

The photovoltaic poverty alleviation (PVPA) project in China integrates clean energy production into poverty reduction, gaining significant academic interest. However, a systematic assessment of the spatial distribution and emission reduction performance of existing PVPA power plants at the site level remains deficient. Based on comprehensive data collection and spatial visualization of all PVPA plants, we revealed spatial distribution patterns and estimates potential environmental benefits for each plant. During 2013–2019, the construction of approximately 180-thansand PVPA plants in China had resulted in an overall installed capacity of 21.77 GW. The spatial layout showed a decreasing trend from east to west, with distinct clusters at the border areas of different provinces. The cumulative emission reduction benefits amounted to 636 million tons, with large-scale solar photovoltaics and the North China Grid contributing significantly. Emission reduction efficiency per unit of installed capacity in the north was higher than the south. This study provides an initial site-level assessment of the PVPA program in China and emphasizes the need for integrating economic feasibility and environmental benefits, along with developing complementary policies and exploring innovative “photovoltaic plus” channels tailored to local contexts in the future.

Bootstrapping outperforms community‐weighted approaches for estimating the shapes of phenotypic distributions

Abstract Estimating phenotypic distributions of populations and communities is central to many questions in ecology and evolution. These distributions can be characterized by their moments (mean, variance, skewness and kurtosis) or diversity metrics (e.g. functional richness). Typically, such moments and metrics are calculated using community‐weighted approaches (e.g. abundance‐weighted mean). We propose an alternative bootstrapping approach that allows flexibility in trait sampling and explicit incorporation of intraspecific variation, and show that this approach significantly improves estimation while allowing us to quantify uncertainty. We assess the performance of different approaches for estimating the moments of trait distributions across various sampling scenarios, taxa and datasets by comparing estimates derived from simulated samples with the true values calculated from full datasets. Simulations differ in sampling intensity (individuals per species), sampling biases (abundance, size), trait data source (local vs. global) and estimation method (two types of community‐weighting, two types of bootstrapping). We introduce the traitstrap R package, which contains a modular and extensible set of bootstrapping and weighted‐averaging functions that use community composition and trait data to estimate the moments of community trait distributions with their uncertainty. Importantly, the first function in the workflow, trait_fill, allows the user to specify hierarchical structures (e.g. plot within site, experiment vs. control, species within genus) to assign trait values to each taxon in each community sample. Across all taxa, simulations and metrics, bootstrapping approaches were more accurate and less biased than community‐weighted approaches. With bootstrapping, a sample size of 9 or more measurements per species per trait generally included the true mean within the 95% CI. It reduced average percent errors by 26%–74% relative to community‐weighting. Random sampling across all species outperformed both size‐ and abundance‐biased sampling. Our results suggest randomly sampling ~9 individuals per sampling unit and species, covering all species in the community and analysing the data using nonparametric bootstrapping generally enable reliable inference on trait distributions, including the central moments, of communities. By providing better estimates of community trait distributions, bootstrapping approaches can improve our ability to link traits to both the processes that generate them and their effects on ecosystems.

Want to Describe and Share Biodiversity Inventory and Monitoring Data? The Humboldt Extension for Ecological Inventories Can Help!

Access to high-quality ecological data is critical to assessing and modeling biodiversity and its changes through space and time. The Darwin Core standard has proven to be immensely helpful in sharing species occurrence data (see Wieczorek et al. 2012, Global Biodiversity Information Facility, GBIF) and promoting biodiversity research following the FAIR principles of findability, accessibility, interoperability and reusability (Wilkinson et al. 2016). However, it is limited in its ability to fully accommodate inventory data (i.e., linked records of multiple taxa at a specific place and time). Information about the inventory processes is often either unreported or described in an unstructured manner, limiting its potential re-use for larger-scale analyses. Two key aspects that are not captured in a structured manner yet are: i) information about the species that were not detected during an inventory, and ii) ancillary information about sampling effort and completeness. Non-detections (i.e., reported counts of zero) potentially enable more accurate and precise estimates of distribution, abundance, and changes in abundance. This becomes possible when variation in effort is used to estimate the likelihood that a non-detection represents a true absence of that taxon during the inventory. Currently, ecological inventory data, when shared at all, are typically discoverable through dataset catalogs (e.g., governmental data repositories) and supplementary materials to publications. With few exceptions, indexing of such data with the detail and structure needed has not been attempted at broad temporal and spatial scales, despite the potentially high value resulting from making inventory data more readily accessible. To address these limitations in documenting inventory data using the Darwin Core, Guralnick et al. (2018) proposed the Humboldt Core. Subsequent discussions within the biodiversity standards community made it clear that greater integration could be achieved by creating an extension of the Darwin Core, rather than developing a new standard in isolation. Extension design work began in 2021 and progress has been reported by Brenton (2021) and Sica et al. (2022). Over the last year the Humboldt Extension Task Group has sought advice from data providers and aggregators and updated its vocabulary terms. A challenging aspect has been creating terminology for the parent-child relationships (see Properties of Hierarchical Events) needed to describe surveys that may be as simple as a collection of checklists (one level of hierarchy) or as complex as species records from traps within plots along transects across habitats over multiple years (at least four levels of hierarchy). The Task Group has committed to completing a User Guide for the Humboldt Extension. Group members who contributed to the Darwin Core (Darwin Core Task Group 2009) and the Vocabulary Maintenance Specification (Vocabulary Maintenance Specification Task Group 2017) have provided valuable expertise on term refinement and process. Through ratification of the Humboldt Extension as a Darwin Core Event extension, we expect to provide the community with a usable solution, tied to well-established data publication mechanisms, for sharing and using inventory data. This effort promises to overcome a key bottleneck in the sharing of critically important ecological data, enhancing data discoverability, interoperability and re-use while lowering reporting burden and data and metadata heterogeneity. Global data aggregation initiatives, such as GBIF, will benefit from this development as they develop their data models and the range of standards and extensions they support. We anticipate that the Humboldt Extension will be attractive both to data publishers and data users, by facilitating the representation and indexing of data in richer, more meaningful ways. Despite the data-intensive nature of fundamental ecological research and applied monitoring for management and policy, ecological data have remained as one of the FAIR data frontiers. We anticipate that the Humboldt Extension will address most data exchange needs of all professional communities involved.

Quantifying the uncertainty of partitions for infinite mixture models

Bayesian clustering models, such as Dirichlet process mixture models (DPMMs), are sophisticated flexible models. They induce a posterior distribution on the set of all partitions of a set of observations. Analysing this posterior distribution is of great interest, but it comes with several challenges. First of all, the number of partitions is overwhelmingly large even for moderate values of the number of observations. Consequently the sample space of the posterior distribution of the partitions is not explored well by MCMC samplers. Second, due to the complexity of representing the uncertainty of partitions, usually only maximum a posteriori estimates of the posterior distribution of partitions are provided and discussed in the literature. In this paper we propose a numerical and graphical method for quantifying the uncertainty of the clusters of a given partition of the data and we suggest how this tool can be used to learn about the partition uncertainty.

PHYSICAL AND MATHEMATICAL MODELING OF THE DISTRIBUTION OF LOAD FORCES ON THE FEMINAL COMPONENT OF AN ENDOPROSTHESIS OF THE HIP JOINT UNDER REAL CONDITIONS

Determination of the factors causing the development of aseptic instability of endoprosthesis components is one of the main tasks of modern traumatology and orthopedics. It is important from a scientific and medical point of view to carry out physical and mathematical modeling of the distribution of load forces and their moments on the femoral component of a hip endoprosthesis. The purpose of the study: to conduct a physical and mathematical modeling of the distribution of load forces on the femoral component of a hip endoprosthesis under real conditions of incomplete axially symmetrical contact of the femoral component of the endoprosthesis and the femur, when the surface of the lower end of the endoprosthesis is not in contact with the surface of the bone. In the work, mathematical modeling of the distribution of point load forces and their moments on the contact surface between the femoral endoprosthesis stem and the bone marrow canal of the femur in real conditions is carried out. For qualitative estimates of point distributions of the load force, based on the analysis of previous results, an estimated empirical formula was obtained for these distributions: g(λ,λ1)=Р(λ,λ1) / 140λ7/2, in which the pressure Р(λ,λ1) is taken in kilograms per square centimeter (kg/cm2), and the point force g(λ,λ1) is in kilograms (kg). It was determined that the best, from the point of view of minimizing the harmful mechanical impact of the prosthesis on the femur, is the situation when the length of the prosthesis stem is not less than half the length of the femur (λ≥0.5). In this case, the values of the point load forces do not exceed 0.1 kg, at least for the length of the area of real contact, which is not less than half the length of the prosthesis stem (λ1≥0.5λ). It has been proven that the use of a prosthesis stem that is less than a third of the length of the femur is not advisable. Since already at the length of the prosthesis stem, which is 30 % of the length of the femur (λ=0.3), point loads increase rapidly and can reach from 0.55 to 1.5 kg depending on the length of the contact area. Such point loads are undesirable for the femur in the area of contact with the prosthesis in terms of the integrity of the femur.

The fuzzy Kullback–Leibler divergence for estimating parameters of the probability distribution in fuzzy data: an application to classifying Vietnamese Herb Leaves

In this paper, we address the challenge of estimating probability distributions which are typically represented by parameter-based values. However, this estimation is prone to errors and does not comprehensively capture the nature of real-world data. Additionally, real-world data often follows a mixed form of probability distributions, where sub-datasets may contain incomplete information. To enhance flexibility, especially in classification problems, we propose a new method for describing parameters estimated through Bayesian statistics. Our method introduces fuzzy parameters and assesses the similarity between probability distributions using the fuzzy extended Kullback–Leibler divergence. We demonstrate the practical application of our approach in Vietnamese Herb Leaves classification. By incorporating fuzzy parameters and leveraging Bayesian statistics, our method provides more robust estimations of probability distributions and enables improved flexibility in classification tasks.

A zero-shot learning boosting framework via concept-constrained clustering

Zero-shot learning (ZSL) aims to recognize novel classes that have no labeled samples during the training phase, which leads to the domain shift problem. In reality, there exists a large number of compounded unlabeled samples. Therefore, it is crucial to accurately estimate the data distribution of these compounded unlabeled samples and improve the performance of ZSL. This paper proposes a zero-shot learning boosting framework. Specifically, ZSL is transformed into a co-training problem between the data distribution estimation of the unlabeled samples and ZSL. The data distribution estimation is modeled as concept-constrained clustering. Furthermore, we design an alternative optimization strategy to realize mutual guidance between the two processes. Finally, systematic experiments verify the effectiveness of the proposed concept-constrained clustering for alleviating the domain shift problem in ZSL and the universality of the proposed framework for boosting different base ZSL models.

Fully Unsupervised Defect Clustering using Adversarial Autoencoder

Identifying defect types and developing proper maintenance strategies is a major concern in modern industry. Most conventional studies have been conducted primarily based on a supervised learning scheme. However, supervised learning has a critical limitation in that it requires labeled data, which is difficult and expensive to obtain in real-world industry. Considering that there are many industries that do not perform post investigations on the defects, fully unsupervised learning methods, which do not exploit any information such as label data or the number of types, need to be developed. Accordingly, in this study, we propose a fully unsupervised defect clustering method that does not exploit any information other than the data itself. The proposed method consists of two major components. The first is dimensionality reduction into latent space via adversarial autoencoder, and the second is a Bayesian mixture model for distribution estimation in latent space. The experiments on a rolling-element-bearing dataset validate the effectiveness of our method. Specifically, our method performs defect clustering without any information other than the data itself, which is promising for real industrial applications.

Distributed Estimation With Cross-Verification Under False Data-Injection Attacks.

Under false data-injection (FDI) attacks, the data of some agents are tampered with by the FDI attackers, which causes that the distributed algorithm cannot estimate the ideal unknown parameter. Due to the concealment of the malicious data tampered with by the FDI attacks, many detection algorithms against FDI attacks often have poor detection results or low detection efficiencies. To solve these problems, a conveniently distributed diffusion least-mean-square (DLMS) algorithm with cross-verification (CV) is proposed against FDI attacks. The proposed DLMS with CV (DLMS-CV) algorithm is comprised of two subsystems: one subsystem provides a detection test of agents based on the CV mechanism, while the other provides a secure distribution estimation. In the CV mechanism, a smoothness strategy is introduced, which can improve the detection performance. The convergence performance of the proposed algorithm is analyzed, and then the design method of the adaptive threshold is also formulated. In particular, the probabilities of missing alarm and false alarm are examined, and they decay exponentially to zero under sufficiently small step size. Finally, simulation experiments are provided to illustrate the effectiveness and simplicity of the proposed DLMS-CV algorithm in comparison to other algorithms against FDI attacks.

Descriptor-Free Deep Learning QSAR Model for the Fraction Unbound in Human Plasma.

Chemical-specific parameters are either measured in vitro or estimated using quantitative structure-activity relationship (QSAR) models. The existing body of QSAR work relies on extracting a set of descriptors or fingerprints, subset selection, and training a machine learning model. In this work, we used a state-of-the-art natural language processing model, Bidirectional Encoder Representations from Transformers, which allowed us to circumvent the need for calculation of these chemical descriptors. In this approach, simplified molecular-input line-entry system (SMILES) strings were embedded in a high-dimensional space using a two-stage training approach. The model was first pre-trained on a masked SMILES token task and then fine-tuned on a QSAR prediction task. The pre-training task learned meaningful high-dimensional embeddings based upon the relationships between the chemical tokens in the SMILES strings derived from the "in-stock" portion of the ZINC 15 dataset─a large dataset of commercially available chemicals. The fine-tuning task then perturbed the pre-trained embeddings to facilitate prediction of a specific QSAR endpoint of interest. The power of this model stems from the ability to reuse the pre-trained model for multiple different fine-tuning tasks, reducing the computational burden of developing multiple models for different endpoints. We used our framework to develop a predictive model for fraction unbound in human plasma (fu,p). This approach is flexible, requires minimum domain expertise, and can be generalized for other parameters of interest for rapid and accurate estimation of absorption, distribution, metabolism, excretion, and toxicity.

Assessing spatiotemporally varied ecohydrological effects of apple orchards based on regional-scale estimation of tree distribution and ages

Assessing spatiotemporally varied ecohydrological effects of apple orchards based on regional-scale estimation of tree distribution and ages