Find research datasets worth reusing

opencc-zeroAug 2021View details →

Data-driven predictions of summertime visits to lakes across 17 US states

Using a dataset of more than 51,000 US lakes, we estimated the relationship between summertime lake visits, lake water quality, landscape features, and other amenities, where visitation was estimated using counts of geolocated photographs. Given the size and complexity of our dataset, we used a combination of machine learning techniques, imputation techniques, and a Poisson count model to estimate these relationships. We found that every additional meter of average summer-time Secchi depth was associated with at least 7% more summer-time lake visits, all else equal. Second, we found that lake amenities, such as beaches, boat launches, and public toilets, were more powerful predictors of visits than water quality. Third, we found that visits to a lake were strongly influenced by the lake's accessibility and its distance to nearby lakes and the amenities the nearby lakes offered. Our research highlights the need for 1) a better understanding of how representative social media data are of actual recreational behavior, 2) the development of best practices to account for non-random patterns in missing natural feature data, and 3) a better understanding of the potential reverse causality in the lake visit-water quality relationships.

Marine Heatwave Prediction

These datasets will be used in the <a href="https://2021noaaaiworkshop.sched.com/">3rd NOAA AI Workshop on Leveraging AI in Environmental Sciences</a> Hackathon for marine heatwave prediction occurring Tuesday–Friday, September 7–10. The goals of this hackathon are to develop innovative ways to understand the relationships between air-sea interactions and the presence/absence of marine heatwave events using historical climatology data.

opencc-by-4.0Mar 2007View details →

Fig. 1 in Environmental factors predicting fish community structure in two neotropical rivers in Brazil

Fig. 1. The Iguatemi River basin, showing the sampling sites in the Jogui and Iguatemi rivers.

Novel Datasets for Evaluating Song Popularity Prediction Tasks

We present two novel datasets for hit song prediction (HSP): HSP-S and HSP-L. They are substantially larger than currently available dataset with respect to the number of features contained. Both datasets provide high- and low-level audio features stemming from <a href="https://acousticbrainz.org/data">AcousticBrainz</a> and short representative MP3 samples. Further, we include listener- and play-counts gathered from <a href="https://www.last.fm">last.fm</a> for both datasets. The larger dataset, HSP-L, contains 73,482 songs with audio features, listener- and play-counts, making it substantially larger than previous datasets. In addition, we provide the release year information for 65,575 songs as provided by the <a href="https://millionsongdataset.com">Million Song Dataset</a>. The smaller dataset, HSP-S, contains 7,736 songs, listener- and play-counts as well as <a href="https://www.officialcharts.com/charts/billboard-hot-100-chart/">Billboard Hot 100</a> data for 50% of the songs, and release year information is available for 7,449 songs.

Why cannot long-term cascade be predicted? Exploring temporal dynamics in information diffusion processes

Predicting information cascade plays a crucial role in various applications such as advertising campaigns, emergency management, and infodemic controlling. However, predicting the scale of an information cascade in a long-term could be difficult. In this study, we take Weibo, a Twitter-like online social platform, as an example, exhaustively extract predictive features from the data, and use a conventional machine learning algorithm to predict the information cascade scales. Specifically, we compare the predictive power (and the loss of it) of different categories of features in short-term and long-term prediction tasks. Among the features that describe the follower-followee network, retweet network, tweet content, and early diffusion dynamics, we find that early diffusion dynamics are the most predictive ones in short-term prediction tasks but lose most of their predictive power in long-term tasks. In-depth analyses reveal two possible causes of such failure: the bursty nature of information diffusion and feature temporal drift over time. Our findings further enhance the comprehension to information diffusion process and may assist in the control of such process.

Supplementary datasets for the paper of "Multi-resBind: a residual network-based multi-label classifier for in vivo RNA binding prediction and preference visualization"

There are two eCLIP datasets (cell lines of K562 and HepG2). The eCLIP datasets were then divided into five categories for each cell line: low, medium 1, medium 2, high 1 and high 2 with peaks of >1,000 but <2,000, >2,000 but <4,000, >4,000 but <7,000, >7,000 but <10,000 and >10,000, respectively.

openmit-licenseSep 2021View details →

kipoi test prediction for APARENT/site_probabilities

During nightly tests, this file will be downloaded and compared against the predictions made by the kipoi model

Data and R code from: Nature calls: intelligence and natural foraging style predict welfare problems in captive parrots

Around half of all parrots (a highly threatened order) live in captivity. Here, some species thrive. Others, however, breed poorly or display stereotypic behaviours indicating stress. Using data on the prevalence of three types of stereotypic behaviour in pet (50 species; 1,378 individuals) and aviculture hatch rates (115 species; 10,255 breeding pairs), we applied Phylogenetic Comparative Methods (PCMs) to test hypothesised causes of this variation (relating to species' rarity and constraints on natural behaviour). In the first empirical evidence that high intelligence increases vulnerability to poor captive welfare, species with large relative brain sizes were found to be most at risk of oral and whole-body stereotypic behaviour. This suggests that if they are to be kept in private homes, such parrots must be offered substantially more cognitive stimulation. Self-harming behaviours involving feather damage were predicted by naturally relying on food items that require substantial handling, highlighting inadequacies in captive diets (often highly processed); while relatively low hatch rates in aviculture were predicted by small captive population sizes, potentially due to genetic bottlenecks, inbreeding, and/or low availability of compatible mates. These novel findings should help advance captive parrot husbandry, and inspire further research applying PCMs to understand and improve animal welfare.

Input Data for "Protein Function Prediction for newly sequenced organisms"

The input sequence files in FASTA format and the detailed list of all organisms excluded when testing each specific bacterium.

Saprotrophic fungal diversity predicts ectomycorrhizal fungal diversity along the timberline in the framework of island biogeography theory

In the context of a timberline tree species (Betula ermanii) as "virtual island", we surveyed ectomycorrhizal (EcM) fungal diversity along a 430-m vertical gradient on the top of Changbai Mountain, China, sampling fine roots and neighboring soils of B. ermanii. Besides elevation, soil properties and plant functional traits, endophytic and saprotrophic fungal diversity were assessed as candidate predictors to construct integrative models. EcM fungal diversity decreased with increasing elevation, and exhibited positive diversity to diameter at breast height and negative diversity to distance from forest edge relationships in both roots and soils. Integrative models further showed that saprotrophic fungal diversity was the strongest predictor of EcM fungal diversity, directly enhancing EcM fungal diversity in roots and soils. All the metadata were stored here for use.

Antibody structure prediction method benchmark results

Antibody Fv structures generated for benchmarking recent antibody structure prediction methods in "Antibody structure prediction using interpretable deep learning."  All successfully predicted targets are provided for each method, although only those where every method succeeded are included in accuracy metrics.  Each DeepAb prediction includes 50 generated decoys from which the final structure was selected.

Combined tumor and immune signals from genomes or transcriptomes predict outcomes of checkpoint inhibition in melanoma

Code and data for manuscript "Combined tumor and immune signals from genomes or transcriptomes predict outcomes of checkpoint inhibition in melanoma"

opencc-by-4.0May 2021View details →

Supplemental Material to Journal Article "Probabilistic Approach for the Fatigue Strength Prediction of Polymers"

This set supplements the figure and table data to the article "Probabilistic Approach for the Fatigue Strength Prediction of Polymers", DOI: <a href="https://doi.org/10.2514/1.J060444">https://doi.org/10.2514/1.J060444</a> The Python code to reproduce the results shown in the article is released as v1.0 of the MAMO toolbox for material models DOI: The latest code development can be cloned at <a href="https://github.com/mrosemeier/mamo">https://github.com/mrosemeier/mamo</a>.

Predicting species abundance by implementing the ecological niche theory

Species are not uniformly distributed across the landscape. For every species, there should be few favoured sites where abundance is high and many other sites of lower suitability where abundance is low. Consequently, local abundance could be thought of as a natural expression of species response to local conditions. The correlation between abundance and environmental suitability has been well documented, and a recent meta-analysis has suggested that this relationship could be a generality. Despite the importance and potential implication of the abundance-suitability relationship, its predictive power for meaningful extrapolations has been surprisingly poorly explored. In this study, we showed how a highly predictable trend can be extracted from the abundance-suitability relationship, accurately predicting the variation in species abundance at a high spatial resolution. We produced high-quality environmental suitability estimations for 50 endemic species to the Australian Wet Tropics. Environmental suitability derived from species distribution models was related to observed abundance estimated using data from 29 years of uninterrupted monitoring effort. We used the fitted relationship to accurately predict abundance at a fine scale across the species range. Our results showed that the abundance-suitability relationship was strong for endemic species in the Australian Wet Tropics. The predictive power of our models was high, explaining, on average, 55% of the deviance across taxa. Despite interspecific variation in the strength of the abundance-suitability relationship associated with potential intrinsic estimation biases, our approach provides a powerful tool for predicting abundance across the species range at a fine scale. The potential for robust abundance predictions from occurrence-based species distribution models shown in this study are numerous, and it could have a significant impact in enhancing species conservation and management decisions.

opencc-by-4.0May 2021View details →

Dataset - seismic data from central-western Italy used in the paper on rapid prediction of ground motion using a Convolutional Neural Network

The dataset published here is the central-western Italy dataset used in the paper "Transfer learning: Improving neural network based prediction of earthquake ground shaking for an area with insufficient training data" (<a href="https://arxiv.org/abs/2105.05075">https://arxiv.org/abs/2105.05075</a>). The code for the paper is available at <a href="https://github.com/djozinovi/TLpredIM">https://github.com/djozinovi/TLpredIM</a>. The abstract of the paper: <blockquote> In a recent study (Jozinović et al, 2020) we showed that convolutional neural networks (CNNs) applied to network seismic traces can be used for rapid prediction of earthquake peak ground motion intensity measures (IMs) at distant stations using only recordings from stations near the epicenter. The predictions are made without any previous knowledge concerning the earthquake location and magnitude. This approach differs from the standard procedure adopted by earthquake early warning systems (EEWSs) that rely on location and magnitude information. In the previous study, we used 10 s, raw, multistation waveforms for the 2016 earthquake sequence in central Italy for 915 events (CI dataset). The CI dataset has a large number of spatially concentrated earthquakes and a dense station network. In this work, we applied the CNN model to an area around area near Pisa, Italy. In our initial application of the technique, we used a dataset consisting of 266 earthquakes recorded by 39 stations. We found that the CNN model trained using this smaller dataset performed worse compared to the results presented in the original study by Jozinović et al. (2020). To counter the lack of data, we adopted transfer learning (TL) using two approaches: first, by using a pre-trained model built on the CI dataset and, next, by using a pre-trained model built on a different (seismological) problem that has a larger dataset available for training. We show that the use of TL improves the results in terms of outliers, bias, and variability of the residuals between predicted and true IMs values. We also demonstrate that adding knowledge of station positions as an additional layer in the neural network improves the results. The possible use for EEW is demonstrated by the times for the warnings that would be received at the station PII. </blockquote>

opencc-by-4.0Oct 2021View details →

Excel sheet to predict Multi-Hydro green roof module's behaviour

To prevent spending long computational times running a lot of simulations using Multi-Hydro it is possible to use this excel sheet to predict the its green roof module's behaviour. It is usefull when some parameters need to be calibrated or estimated before the final simulation.

opencc-by-4.0Dec 2020View details →

Data and R-Scripts: Predicting hotspots for invasive species introduction in Europe

The .rar file comprises all data and R-scripts needed to replicate our study entitled "Predicting Hotspots for Invasive Species Introduction in Europe" published in Environmental Research Letters. The folder data holds all input data as well as the final datasets used for training the algorithms, in the subfolder A_ML_ready_datasets. The folder descriptives provides tables with descriptive statistics. The folder figures provides files for all figures displayed in the manuscript and the supplementary material as well as visualizations of descriptive statistics for all background approaches in the corresponding subfolders. The folder results holds all generated results. The folder scripts provides all R-scripts used for intermediate computations. The master and master_results scripts coordinate all computations and the generation of results, respectively. Notably, various spatial layers were used to extract point-values of features which subsequently were used to estimate the models and generate the predictions. Here, we only upload the extracted point-values in the data folder. If you are interested in using any of the raw spatial layers, please refer to section 2.1.3. of our publication to find the corresponding references. Alternatively, feel free to reach out to me and I will direct you to the original databases and/or send you the raw spatial layers.

opencc-zeroOct 2021View details →

The telomere regulatory gene POT1 responds to stress and predicts performance in nature: implications for telomeres and life history evolution

Telomeres are emerging as correlates of fitness-related traits and may be important mediators of ecologically relevant variation in life history strategies. Growing evidence suggests that telomere dynamics can be more predictive of performance than length itself, but very little work considers how telomere regulatory mechanisms respond to environmental challenges or influence performance in nature. Here, we combine observational and experimental datasets from free-living tree swallows (Tachycineta bicolor) to assess how performance is predicted by the telomere regulatory gene POT1, which encodes a shelterin protein that sterically blocks telomerase from repairing the telomere. First, we show that lower POT1 gene expression was associated with higher female quality, i.e. earlier breeding and heavier body mass. We next challenged mothers with an immune stressor (lipopolysaccharide injection) that led to 'sickness' in mothers and 24h of food restriction in their offspring. While POT1 did not respond to maternal injection, females with lower constitutive POT1 gene expression were better able to maintain feeding rates following treatment. Maternal injection also generated a one-day stressor for chicks, which responded with lower POT1 gene expression and elongated telomeres. Other putatively stress-responsive mechanisms (i.e. glucocorticoids, antioxidants) showed marginal responses in stress-exposed chicks. Model comparisons indicated that POT1 mRNA abundance was a largely better predictor of performance than telomere dynamics, indicating that telomere regulators may be powerful modulators of variation in life history strategies.

openother-openOct 2021View details →

Data on Predictive Resilience of Interdependent Water and Transportation Infrastructures: A Sociotechnical Approach

Interdependent critical infrastructures are vulnerable to the cascading effect of failures resulting from random failures and natural disasters. The data provided in this work is the processed data used for the proposed resilience assessment framework for interdependent water-transportation networks dealing with both types of failure \citep{1}. The case study is the interconnected networks of water and transportation in Tampa, Florida. The data for the random failure is obtained from the developed algorithmic framework and the land use and social vulnerability data provided by U.S. Census datasets. We then used a subset of this produced data to construct the predictive models for the network resilience to random failures. As for the natural disaster scenario, we focused on hurricane Irma in 2017 as it directly affected the focused region in Florida. We used the specific guidelines and the raw flooding data for this hurricane provided by FEMA to estimate the standing water for each geographical area (polygons) and the associated network components. We labeled the areas as failed and undamaged based on the estimated water levels. Finally, we used this data for developing a geospatial Geographical Weighted Regression (GWR) model to predict the resilience in each polygon. We present the final dataset for water and transportation networks to facilitate reusability for any future resilience study in the selected urban area.