Share on FacebookShare on X (formerly Twitter)Share on PinterestShare on LinkedInPANAMA CITY
Anastacio this week as she received her Golden Apple Award
Teaching at Patronis Elementary School for 30 years
she said she drives in everyday from Destin because it is the best
She said she enjoys teaching her 2nd graders because
“They’re like little sponges and they take in everything that I can give them and they’re just a lot of fun.”
She describes everyday as being different with many different ways to approach the day
you will also have fun and your day flies by.”
Anastacio explains the benefits of staying in one place and how she would like to thank the wonderful students
“For teaching this many years you get to know that everybody has different skills
Everybody brings different things to the table
it’s just been great being in the same place for so many years.”
To stay up to date on all the latest news as it develops, follow WJHG on Facebook, Instagram and X (Twitter)
Have a news tip or see an error that needs correction? Email news@wjhg.com
Please include the article’s headline in your message
Keep up with all the biggest headlines on the WJHG News app, and check out what’s happening outside using the WJHG Weather app
a paraprofessional at Patronis Elementary School in Panama City Beach
was called to the principal's office on April 3
the tall Brazilian with big-big energy was surprised by several colleagues
they presented him a bouquet and certificate of statewide achievement
Bay District Schools was notified by the Florida Department of Education that their trilingual language assistant was recognized as one of the state's top five school-related employees of the year
In January, Anastacio won the “Support Employee of the Year” award during the 2024-25 Excellence in Education Awards Ceremony held at Bay High School
As a key figure in Patronis Elementary's international community
Anastacio bridges language barriers for students
ensuring every child has the support they need to succeed
said Anastacio represents everything about Bay County and the state."He's just all the things that students love," Loyed said
that's 90% of the battle … just get pumped up about the day."
Anastacio said he wasn't expecting the recognition but wanted to thank everyone who trusted him and supported him on the big achievement
The language facilitator will represent Bay County at the state competition in Orlando later this year
Anastacio is currently enrolled at Gulf Coast State College and wants to become an English as a Second Language educator
(That's Portuguese for "Congratulations on all your hard work!")
(WJHG/WECP) - A Bay District Schools’ employee is making a statewide impact
Patronis Elementary School’s language assistant received a surprise announcement—he’s a finalist for Florida’s top school employees
Shock, excitement, and pure joy filled the room as Mr. Eduardo Anastacio learned he is a finalist among Florida’s top five school-related employees of the year for 2025. Just months ago, he was named Bay District Schools’ support employee of the year
and he was just as excited then as he is now
The Florida Department of Education’s program recognizes outstanding educational support personnel
honoring the contributions they make to their schools
This Brazilian native teaches English to 38 foreign students at Patronis Elementary School
making them his biggest motivation to go to school every day
“As a language assistant I have to help kids learn English
and dynamic; are all qualities that represent Mr
“He’s just everything you would want in a support employee or a full-time faculty member
he’s just all the things that students love; he’s personable
he has energy,” said Patronis Elementary School’s principal Brooke Loyed
Anastacio also has a second job and is working towards his teaching certification to soon become a full-time ESOL teacher
“I remember learning English by myself when I was taking a shower
He will now be heading to the state competition to represent Bay District Schools
This website is using a security service to protect itself from online attacks
The action you just performed triggered the security solution
There are several actions that could trigger this block including submitting a certain word or phrase
You can email the site owner to let them know you were blocked
Please include what you were doing when this page came up and the Cloudflare Ray ID found at the bottom of this page
Narcotics Rewards Program: Brought to Justice
the Sinaloa Cartel is largely responsible for the massive influx of fentanyl into the U.S
On March 18, 2023, Soto Vega was arrested in Athens, Greece, on a provisional arrest warrant based on charges in a criminal complaint which were later included in an indictment. He remains in custody pending extradition to the U.S..
On April 4, 2023, a federal grand jury in the Southern District of New York returned an indictment against Anastacio Soto Vega and others charging them with Fentanyl Importation Conspiracy, Possession of Machineguns & Destructive Devices, and Conspiracy to Possess Machineguns & Destructive Devices.
The U.S. Department of State is offering a REWARD OF UP TO $1 MILLION for information leading to the conviction of Anastacio Soto Vega.
If you have information, please contact the DEA via email at ChapitosTips@dea.gov. If you are located outside of the U.S., you may also visit the nearest U.S. Embassy or Consulate. If in the U.S., you can contact the local DEA office in your city.
ALL IDENTITIES ARE KEPT STRICTLY CONFIDENTIAL.
Government officials and employees are not eligible for rewards.
We use cookies to make our website work better and improve your experience. By continuing to use the site, you agree to our privacy policy.
Volume 9 - 2021 | https://doi.org/10.3389/fchem.2021.787194
Despite the increasing number of pharmaceutical companies
less than one percent of initially researched drugs enter the commercial market
virtual screening (VS) has gained much attention due to several advantages
reduced reagent and consumable costs and the performance of selective analyses regarding the affinity between test molecules and pharmacological targets
VS is based mainly on algorithms that apply physical and chemistry principles and quantum mechanics to estimate molecule affinities and conformations
VS has not reached the expected results concerning the improvement of market-approved drugs
comprising less than twenty drugs that have reached this goal to date
may comprise a powerful tool to improve VS results concerning natural products that may be used both simultaneously with standard algorithms or isolated
This review discusses the pros and cons of GNN applied to VS and the future perspectives of this learnable algorithm
which may revolutionize drug discovery if certain obstacles concerning spatial coordinates and adequate datasets
Mathematic modeling comprises a valuable important tool in the development of the pharmacology and chemistry fields since their beginning as formal disciplines (Gaddum, 1953; Atkinson and Lalonde, 2007; Finlay et al., 2020)
most traditional pharmacologists and chemistries are very accustomed in employing math modeling to solve or aid current issues regarding the development of new drugs
The first chemoinformatic assessment was reported by Ray and coworkers (1957)
who employed a new algorithm to detect molecular substructures
Representations of tensors in different orders (dimension
Column 1 lists the elements that compose the sets corresponding to each dimension
The number of sets is in agreement with the number of dimensions except for the dimension 0
as one element only is used to represent a scalar
More elements could have been used in each set for dimension 1 and above
but only two elements were used to reduce the complexity of the figure
comprising the same two elements in all dimensions
Column 2 contains elements “boxed” and spatially arranged in correspondence with the respective data structure
These comprise a row vector of “boxes” for dimension 1
a matrix with rows and columns of “boxes” for dimension 2
Column 3 exhibits an equivalent arrangement as in the preceding column
except that each element is represented with “one-hot” encoding instead of numeric encoding
The “one-hot” representation is usual in neural networks to input information representing a set of features
Each feature is represented by a finite set of possible values of a specific feature attribute
with values ordered as positions in a sequence of values “0” and “1.” As an example
if a specific feature has an attribute summarized as having N possible values
it is represented by a sequence where one numeric value “1” is located in correspondence with the observed value of the attribute and N-1 values “0” are located in the positions corresponding to the other values
composed by a longer sequence chaining the “one-hot” representation of all features
Column 4 of the table presents examples of tensor representations as they are usually coded in computational language (Python coding
At the beginning and ending of the data representation
the number of square brackets must conform to the tensor dimension
Columns 5 and 6 are the nomenclature of each kind of tensors up to rank 4 and its respective dimensions
This review focus on a subtype of deep learning algorithm named graph neural network (GNN), currently one of the most applied. Despite being recent, the use of deep learning algorithms employing GNN may revolutionize the VS field, considered by some authors as the state of the art due to its high accuracy rates (Gaudelet et al., 2021)
This article is mainly directed towards scientists
teachers and students that work with drug discovery or enthusiasts concerning this topic
a more didactic language and clear figures will be employed to clarify this theme and
“recruit” new users of this technology
Structure-based drug design is mainly employed when the structure of target is known through the application of structural methodologies
nuclear magnetic resonance and cryo-electron microscopy
The term “deep” in deep learning indicates a hidden layer or “hidden neurons,” as this network is initially based on neural functioning
and it is important to compare “math neurons” to biological neurons to better understand deep learning impacts in any modern world area
a synapse between the sensitive neuron with an interneuron that synapses onto
(A) Schematic representation of a hand touching a hot pan and receiving a temperature sensory receptor stimulus
(B) The stimulus provokes a signal that will travel through a sensory neuron into the spinal cord
releasing excitatory neurotransmissions when it synapses with an alpha motoneuron that innervates the biceps
The sensory neuron will also synapse with an interneuron
that synapses with an alpha motoneuron that innervates the triceps
although in this case inhibitory neurotransmitters are released
(C) Channels from the TRP family expressed in sensory neurons will open following the thermal stimulus
leading to an entrance of cations such as sodium
resulting in membrane depolarization that could trigger an action potential
(D) Reflex of removing the hand from the heat after touching the pan
It is important to note this movement is performed through the association of inhibitory and excitatory synapses that work as positive and negative weights in artificial neural networks
Comparative model of perceptron and real neuron
A neuron receives both excitatory and inhibitory signals through the release of different neurotransmitters from different biological neural networks that generates postsynaptic excitatory or inhibitory potentials
These potentials can add up with space or time in a continuous manner
If this voltage reaches the potential threshold
it will trigger an AP in the axon implantation cone (hillock)
which will lead to the passage of the signal through the axon and the axon terminals until it synapses with other neurons
This type of signal can be classified as digital
because it occurs (binary 1) or not occurs (binary 0)
depending on the signal reaching the action potential threshold
It is important to point out that the discovery of the action potential mechanism comprised a fantastic association of bench work and modeling performed by Huxley and Hodgkin with seminal papers published after the Second World War comprising a cornerstone of computational neuroscience (Schuetze, 1983; Hausser, 2020). In same period, McCulloch and Pitts (1943) were the first to mathematically model a neural network
Macculloch had an undergraduate degree in psychology and a graduate degree in medicine
both engineers that worked on the Second World War radar defense system
and his team at the Cornell Aeronautical Laboratory developed the perceptron algorithm implemented in an IBM 704
consisting of the first “perceptron computer” to detect visual stimuli
a significant perceptron innovation comprised weights that act as excitatory and inhibitory neurotransmitters chemical synapses in neurons
The nervous system transduces external and internal signal as explained above through receptor potentials
The receptor potential and postsynaptic potentials can be added temporally and spatially
act as an aggregation function in perceptron and deep learning methods
These potentials can trigger an action potential in areas displaying a high number of voltage-gated sodium channels when the potential reaches a certain threshold
the analogic signal become a digital signal
generates different frequencies to “pass along the message.”
binary step function or heaviside function–This function was used in the early perceptron era
It is unsuitable for gradient descent learning methods
since its derivative is 0 along almost the entire domain
(B) Logistic or sigmoid function—This is perhaps the most extensively applied activation function
depending on the value of parameter δ
(C) Hyperbolic tangent (TanH)—This function is used as an alternative to the sigmoid function when a range of output from −1 to 1 is needed
instead of the range from 0 to 1 given by the sigmoid function
(D) Softplus function or rectifier (as the shape of the function resembles the behavior of a rectifier diode)—It is differentiable along the entire domain as the two previous functions
since its derivative exhibits non-zero values along the entire positive domain while sigmoid and hyperbolic tangent present derivatives approaching zero asymptotically for large argument values
A non-zero derivative along a broader range may contribute to speed up the learning process
The derivative of the softplus function is the sigmoid function
(E) Rectified linear unity function (ReLU)—This function is used as a simple linear alternative to the softplus
maintaining the same advantage of a non-zero derivative along the entire positive domain but suffering the problem of a discontinuity in the derivative for argument equal to 0
(F) Leaky ReLU—This function is a “leaky” form of ReLU
The answer of this function in the negative domain comprises an attenuated version of the linear response in positive domain
Depending on the attenuation parameter being a constant or an adjustable value
the function may be also named Parametric ReLU
The non-zero negative domain provides some decrease in the occurrence of dead neurons
the region of ReLU curve where any signal corresponds to zero as a response
where an illustrative example of a prediction using message passing neural network (MPNN) is provided
including graphical (such as 3D structures)
One way of representing a molecule is to use graphs that work as a kind of an abstract version of a molecular modeling kit
The instructions to mount a given molecule model with N atoms labeled by sequential numbering (formally known as nodes) can be given by a table with N rows and N columns
where the existence of a bond (named edge) between each pair of atoms i and j is represented by a value 1 in row i and column j in the given table (named adjacency matrix)
and with every other position in the table presenting value 0
A molecular prediction process scheme employing a message passing neural network (MPNN) as molecule embedder
(A) Ammonia was chosen as a candidate for prediction
(B) Text representation of ammonia as a simplified molecular-input line-entry system (SMILES) sequence and molecular formula
(C) The SMILES sequences are converted to a graph with three edges representing a molecular bond between hydrogen (green) and nitrogen (yellow)
(D) h0 is the feature space for a specific atom
is didactically associated with the corresponding atom number
(E) These features comprise the atom features
which are averaged with the message function
(F) An example of how the first step of the MPNN is computed for nitrogen (n4) is provided
(G) The final step of MPNN is a feature embedding for each atom
which can be summed to be then used by a feed-forward network for final predictions
(H) M is the feature embedding for each atom
consisting in a shape of (N Atoms x E Embedding space)
This (NxE) tensor can be feeded into a more simplistic network architechture
i.e.: Feed Forward to train task-specific models
Summary of main graph neural network architectures
employing schematic neurons to demonstrate the signal passage through different layers
The predicted value named ŷ is obtained in the final layer
This value will be decreased from the real y value
The process is repeated until the smallest possible error is reached through a backpropagation algorithm
It is possible to extrapolate a molecule representation in the same graph-like structure of a social network, where atoms comprise the people and bonds are equivalent to relationships. It is also possible to convert a textual representation into a graph representation. Figure 6 provides an example of this conversion with ammonia
represented as its molecular formula and SMILES (simplified molecular-input line-entry system)
The latter presents some conventions that make molecule interpretation easier by computer programs
The sequence is converted into a graph displaying the aforementioned edge and bond relationship
graphs allow for not only modeling relationships
but also for adding node and edge information
it is possible not only to store how friends are related
but the name of each friend (which would be node-added information) or when the friendship began (edge-added information)
this allows for information on atoms (i.e.
bond type) to be stored alongside the bonds and atoms that represent the molecule itself
information coding concerning atomic features corresponding to one value choice among a set (for example
the kind of the atom in a node being chosen from a list of M possible atoms) may be performed by “one-hot” coding
where the position in the list that corresponds to the specific feature of the atom occupying the node being set to 1 and the remaining M-1 positions
considering a short list of possible atoms as (H
the nitrogen in the ammonia molecule would be represented by (0,0,1,0,0,0,0,0) and each of the hydrogen atoms by (1,0,0,0,0,0,0,0)
Features that can be expressed by integer values may be coded by the integer value itself or also by “one-hot” coding considering a list of all possible values attributable to the investigated feature
considering an entire set of atomic features is obtained by concatenating the codes attributed to all these features in the form of a vector in a defined order
it is usual for these vectors to present a dimension around one hundred or more
The features of the entire molecule are thus
represented by a matrix whose rows corresponds to the feature vectors of the atoms according to the sequential order in which they are labeled
even small molecules may be represented by matrices with hundreds to thousands of values
A neural network layer with a matrix of features like these as input and a number of neurons about the same order as outputs would easily surpass millions of parameters to be adjusted by training
In this context, convolution techniques have been proven useful to enhance molecular features while, at the same time, significantly reduce the dimension of the feature matrix, as depicted in Figure 7
Convolution has been extensively employed in image or language processing and consists in multiplying a small matrix
known as “filter” or “kernel,” by the data matrix (the feature matrix
The dimension of the filter usually comprises few rows and columns and the multiplication takes the same number of elements as the filter from the data matrix
considering the element in the first row and the filter column being aligned with one element in the data matrix in a variable position
for each element in the filter a corresponding element is noted in the data matrix with an equivalent offset of rows and columns regarding the first filter element
The data matrix and filter elements are multiplied position by position and summed (an operation equivalent to the scalar product)
the scalar product is obtained for each alignment position by moving the filter from the first row and first column of the data matrix
known as “stride,” corresponds to a value in the range from 1 to the corresponding filter dimension
(A) Example of an atomic feature matrix and an adjacency matrix for ammonia
j in the adjacency matrix are representations of the connection between atoms i and j
A combination of the feature and adjacency matrix is performed by column wise multiplication
Each atomic bond corresponds to a “page” in the array
A filter (kernel) corresponding to a 3D matrix with the same number of pages as the combined data set feature-adjacency is applied to the matrix through a scalar product
when inserted in a neural network comprise weight parameters for the connections and its values are adjusted by training
defining the filter characteristics in an optimized manner
The offset of the filter with respect to the data set is swept to cover the entire indices range
The result of the convolution operation in this example is a 2D array with four rows and N-2 columns
where N is the number of columns in the feature matrix
(B) Example of a hypothetical neural network used to calculate the value of a property of a given molecule
Several convolution layers are chained to perform the embedding of the data representing the atomic features and the atomic bonding
the result being a 1D vector which is further submitted to a fully connected neural network
The output of this network is the desired parameter
The training algorithm adjusts the parameters of all kernel filters and the weights of the output neural network
until the error of the predictions compared to the training set are minimized
Comparing the dimension of the original data matrix
with to the corresponding dimension of the convoluted matrix
and convolution filter displaying a dimension J rows by K columns and a stride value of Sr and Sc for rows and columns
when dimensionality reduction is not a desired convolution result
a convenient filling with all zero to rows and columns around the data matrix may be performed previously to convolution
known as “padding.” Considering the addition of Pr rows with zeros above and below the data matrix and Pc columns with zeros on the left and right of the data matrix
techniques combining the features matrix with the adjacency matrix forming a multidimensional array followed by use of convolution techniques to obtain smaller arrays or even vectors as a result
a general process usually known as “embedding,” provides a more compact representation joining atomic and bond features
This representation is much more convenient to be used as input in a final fully connected feed forward neural network to obtain the final value of the molecular property of interest being modeled
Chaining groups of graph representation layers and graph convolution layers in a network structure to predict molecular properties has been reported as achieving superior performance in some molecular property predictions (Wang et al., 2019)
Despite having the same single bond to the new hydrogen as it had to the previous hydrogen atom
ammonium displays many different properties compared to ammonia
The addition of a new hydrogen affects the other atoms in the molecule
altering its shape from triangular pyramidal to tetrahedral
This can be construed as the message propagating from one atom and bond (from the nitrogen to the new hydrogen) to the others atoms (the remaining hydrogens)
This message function is one of the learned functions within the MPNN
the purpose of the MPNN is to convert the unstructured data of a graph (which previously comprised a simple text) into a semantic embedding which essentially comprises a tensor assumed to be the best molecule summarization
This summarization in the form of a tensor can be applied to any subsequent task
such as predicting blood-brain barrier permeability
The base assumption is that learning how to best summarize the molecule should simplify any following prediction that employs this summarization
Recently, Wieder et al. performed a literature survey accounting for about 80 different GNN models in 63 publications, which were applied to different fields such as quantum chemistry, physicochemical property predictions, biophysics, biological effects, and synthetic accessibility (Wieder et al., 2020)
This section discussed some recent GNN applications to VS field
Currently, an increasing number of articles describing new frameworks to predict interactions between ligands and proteins is noted (Jin et al., 2021)
these graph-based neural networks are gaining new adaptations and
constantly exhibit better performance than conventional molecular docking programs
Jiang et al. created an accurate model (<96%) to predict drug-target interactions, based on the construction of two graphs: one for the molecule according to its SMILES sequence, and one protein graph built from a contact map of the protein sequence. Subsequently, two GNN extracted the information and were able to predict the affinity of the ligand and the target protein (Jiang et al., 2020)
Furthermore, GNN algorithms can be used to predict EC50, solubility, and molecular properties. They are also able to perform molecular dynamics simulations (Duvenaud et al., 2015; Klicpera et al., 2020)
Although GNN show better results in terms of accuracy than molecular docking methodologies, for example, their applications to VS are still scarce and studies are still recent. This demonstrates that VS is still not able to keep up with the growth rate of the improved GNN models currently being produced. However, this will probably change soon. VS results obtained by GNN are summarized in Table 2
In another study, Hsieh et al. used GNN methodology to discover repurposable drugs to treat COVID-19. Their model was constructed based on the SARS-CoV-2 knowledge graph map, which considers several virus interactions such as baits, host genes, pathways, phenotypes, and drugs. Their work highlighted 22 potential drugs (Hsieh et al., 2020)
Liu et al. performed a VS to discover novel anti-osteoporosis drugs from natural products using a pre-trained self-attentive message passing neural network (P-SAMPNN). Among the five hits selected for in vitro tests, a laudanosine derivative and a codamine derivative exhibited activity at the nanomolar range (i.e., 32 and 68 nM, respectively), suppressing osteoclastogenesis-related genes (Liu et al., 2021)
Best performance values are mostly associated with graph-based models, with few exceptions comprising non-graph models performing better than graph models when applied to specific databases and properties (Jiang et al., 2021)
RF and XGBoost providing the best AUC-ROC values
There may be enough room to find synergism in combinations of graph based on descriptor based models to achieve improved results
stating that GNNs outperform other ML methods such as Multilayer Perceptrons for chemical predictions
regarding SMILES sequences being similar to Natural Language Processing (NLP) sequences
further assessments on evaluating GNNs against Transformer-based DL methods
that could simplify inputs and are better described in NLP applications
these findings indicate that the use of deep learning tools can aid in overcome the long-standing challenges surrounding natural product research
as well as accelerate the drug discovery process
and NF prepared the tables; All authors critically reviewed the manuscript and improved it
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations
Any product that may be evaluated in this article
or claim that may be made by its manufacturer
is not guaranteed or endorsed by the publisher
We thank Oswaldo Cruz Foundation and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) for financial support
We also thank the Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ) for the financial support (“Redes de Pesquisa em Saúde no Estado do Rio de Janeiro” - Grant number: E-26/010.002422/2019)
CrossRef Full Text | Google Scholar
Introduction of Quantitative Methods in Pharmacology and Clinical Pharmacology: A Historical Overview
PubMed Abstract | CrossRef Full Text | Google Scholar
Google Scholar
Data Resources for the Computer-Guided Discovery of Bioactive Natural Products
PubMed Abstract | CrossRef Full Text | Google Scholar
XGraphBoost: Extracting Graph Neural Network-Based Features for a Better Prediction of Molecular Properties
CrossRef Full Text | Google Scholar
Dobrev, D. (2005). A Definition of Artificial Intelligence, New SeriesFasc. 1–2. Mathematica Balkanica 19, 67–74. Available at: http://arxiv.org/abs/1210.1568 (Accessed December 7
Google Scholar
“Convolutional Networks on Graphs for Learning Molecular Fingerprints,” in Advances in Neural Information Processing Systems 28
Google Scholar
Dwivedi, V. P., Joshi, C. K., Laurent, T., Bengio, Y., and Bresson, X. (2020). Benchmarking Graph Neural Networks. arXiv:2003.00982v3. Available at: https://arxiv.org/abs/2003.00982 (Accessed December 20
Google Scholar
Solutio Problematis Ad Geometriam Situs Pertinentis
Academiae Scientiarum Imperialis Petropolitanae 8
Google Scholar
A Brief History of Simulation Neuroscience
PubMed Abstract | CrossRef Full Text | Google Scholar
100 Years of Modelling Ligand-Receptor Binding and Response: A Focus on GPCRs
CrossRef Full Text | Google Scholar
PubMed Abstract | Google Scholar
Garg, V. K., Jegelka, S., and Jaakkola, T. (2020). Generalization and Representational Limits of Graph Neural Networks. arXiv:2002.06157v1. Available at: https://arxiv.org/abs/2002.06157 (Accessed December 20
Google Scholar
Chemoinformatics: Achievements and Challenges
PubMed Abstract | CrossRef Full Text | Google Scholar
Utilizing Graph Machine Learning within Drug Discovery and Development
PubMed Abstract | CrossRef Full Text | Google Scholar
An Open-Source Drug Discovery Platform Enables Ultra-large Virtual Screens
PubMed Abstract | CrossRef Full Text | Google Scholar
Häusser
The Hodgkin-Huxley Theory of the Action Potential
PubMed Abstract | CrossRef Full Text | Google Scholar
Hsieh, K., Wang, Y., Chen, L., Zhao, Z., Savitz, S., Jiang, X., et al. (2020). Drug Repurposing for COVID-19 Using Graph Neural Network with Genetic, Mechanistic, and Epidemiological Validation. arXiv [Preprint|. Available at: https://arxiv.org/abs/2009.10931v1 (Accessed September 27
Google Scholar
Quantitative Structure‐Toxicity Relationship: An “In Silico Study” Using Electrophilicity and Hydrophobicity as Descriptors
CrossRef Full Text | Google Scholar
Could Graph Neural Networks Learn Better Molecular Representation for Drug Discovery
A Comparison Study of Descriptor-Based and Graph-Based Models
CrossRef Full Text | Google Scholar
Drug-target Affinity Prediction Using Graph Neural Network and Contact Maps
CrossRef Full Text | Google Scholar
Application of Deep Learning Methods in Biological Networks
PubMed Abstract | CrossRef Full Text | Google Scholar
Deep Learning for Drug Design: An Artificial Intelligence Paradigm for Drug Discovery in the Big Data Era
PubMed Abstract | CrossRef Full Text | Google Scholar
How Far Can Virtual Screening Take Us in Drug Discovery
PubMed Abstract | CrossRef Full Text | Google Scholar
Kim, H. W., Wang, M., Leber, C. A., and Nothias, L. (2020). NPClassifier: A Deep Neural Structural Classification Tool for Natural Products. ChemRxiv [Preprint]. Available at: https://ChemRxiv.org.10.26434/chemrxiv.12885494.v1 (Accessed September 27
Google Scholar
“Directional Message Passing for Molecular Graphs,” in International Conference on Learning Representations
Google Scholar
“Artificial Intelligence: Definition
and Cases,” in Encyclopedia of Life Support Systems (EOLSS)
Developed under the Auspices of the UNESCO
Google Scholar
PubMed Abstract | CrossRef Full Text | Google Scholar
PubMed Abstract | CrossRef Full Text | Google Scholar
CrossRef Full Text | Google Scholar
Li, J., Cai, D., and He, X. (2017). Learning Graph-Level Representation for Drug Discovery. Available at: https://arxiv.org/abs/1709.03741v2 (Accessed December 7
Google Scholar
Li, R., Wang, S., Zhu, F., and Huang, J. (2018). Adaptive Graph Convolutional Neural Networks. Proc. AAAI Conf. Artif. Intell. 32, 3546–3553. Available at: https://ojs.aaai.org/index.php/AAAI/article/view/11691 (Accessed December 7
Google Scholar
Lichtarge, O., Sowa, M. E., and Philippi, A. (2002). Evolutionary Traces of Functional Surfaces along G Protein Signaling Pathway. Methods Enzymol. 344, 536–556. Available at: http://www.ncbi.nlm.nih.gov/pubmed/11771409
doi:10.1016/s0076-6879(02)44739-8 (Accessed December 7
PubMed Abstract | CrossRef Full Text | Google Scholar
Predicting Drug-Target Interaction Using a Novel Graph Neural Network with 3D Structure-Embedded Graph Representation
PubMed Abstract | CrossRef Full Text | Google Scholar
History of Solid Organ Transplantation and Organ Donation
PubMed Abstract | CrossRef Full Text | Google Scholar
A Strategy on Selecting Performance Metrics for Classifier Evaluation
CrossRef Full Text | Google Scholar
Deep Learning Enables Discovery of Highly Potent Anti-osteoporosis Natural Products
CrossRef Full Text | Google Scholar
Structure-based Virtual Screening: From Classical to Artificial Intelligence
PubMed Abstract | CrossRef Full Text | Google Scholar
Advances with Support Vector Machines for Novel Drug Discovery
PubMed Abstract | CrossRef Full Text | Google Scholar
A Logical Calculus of the Ideas Immanent in Nervous Activity
CrossRef Full Text | Google Scholar
Using Domain-specific Fingerprints Generated Through Neural Networks to Enhance Ligand-Based Virtual Screening
CrossRef Full Text | Google Scholar
Structural Mechanism of Heat-Induced Opening of a Temperature-Sensitive TRP Channel
PubMed Abstract | CrossRef Full Text | Google Scholar
Origin and Evolution of High Throughput Screening
CrossRef Full Text | Google Scholar
NuBBEDB: an Updated Database to Uncover Chemical and Biological Information from Brazilian Biodiversity
PubMed Abstract | CrossRef Full Text | Google Scholar
A Convolutional Neural Network-Based Approach for the Rapid Annotation of Molecularly Diverse Natural Products
PubMed Abstract | CrossRef Full Text | Google Scholar
Recent Applications of Deep Learning and Machine Intelligence on In Silico Drug Discovery: Methods
CrossRef Full Text | Google Scholar
“Using Deep Siamese Neural Networks to Speed up Natural Products Research,” in International Conference on Learning Representations
Google Scholar
CrossRef Full Text | Google Scholar
Learning Representations by Back-Propagating Errors
CrossRef Full Text | Google Scholar
Prediction of Pharmacological Activities from Chemical Structures with Graph Convolutional Neural Networks
PubMed Abstract | CrossRef Full Text | Google Scholar
How AI and Neuroscience Drive Each Other Forwards
PubMed Abstract | CrossRef Full Text | Google Scholar
Deep Learning in Neural Networks: An Overview
PubMed Abstract | CrossRef Full Text | Google Scholar
CrossRef Full Text | Google Scholar
Shang, C., Liu, Q., Chen, K.-S., Sun, J., Lu, J., Yi, J., et al. (2018). Edge Attention-Based Multi-Relational Graph Convolutional Networks. Available at: https://github.com/Luckick/EAGCN (Accessed December 7
Google Scholar
Stähelin
The History of Cyclosporin A (Sandimmune) Revisited: Another point of View
PubMed Abstract | CrossRef Full Text | Google Scholar
A Deep Learning Approach to Antibiotic Discovery
PubMed Abstract | CrossRef Full Text | Google Scholar
Thompson, N. C., Greenewald, K., Lee, K., and Manso, G. F. (2020). The Computational Limits of Deep Learning. arXiv:2007.05558v1. Available at: https://arxiv.org/pdf/2007.05558.pdf (Accessed September 27
Google Scholar
Graph Convolutional Neural Networks for Predicting Drug-Target Interactions
PubMed Abstract | CrossRef Full Text | Google Scholar
Velicković
“Graph Attention Networks,” in Sixth International Conference on Learning Representations
Google Scholar
Molecule Property Prediction Based on Spatial Graph Embedding
PubMed Abstract | CrossRef Full Text | Google Scholar
Toward Heterogeneous Information Fusion: Bipartite Graph Convolutional Networks for In Silico Drug Repurposing
PubMed Abstract | CrossRef Full Text | Google Scholar
A Compact Review of Molecular Property Prediction with Graph Neural Networks
PubMed Abstract | CrossRef Full Text | Google Scholar
A Comprehensive Survey on Graph Neural Networks
PubMed Abstract | CrossRef Full Text | Google Scholar
MoleculeNet: a Benchmark for Molecular Machine Learning
PubMed Abstract | CrossRef Full Text | Google Scholar
Pushing the Boundaries of Molecular Representation for Drug Discovery with the Graph Attention Mechanism
PubMed Abstract | CrossRef Full Text | Google Scholar
Analyzing Learned Molecular Representations for Property Prediction
CrossRef Full Text | Google Scholar
A Deep Learning-Based Approach for Identifying the Medicinal Uses of Plant-Derived Natural Compounds
PubMed Abstract | CrossRef Full Text | Google Scholar
“On Definition of Deep Learning,” in World Automation Congress (WAC) Stevenson
CrossRef Full Text | Google Scholar
PubMed Abstract | CrossRef Full Text | Google Scholar
A Novel Graph Neural Network Methodology to Investigate Dihydroorotate Dehydrogenase Inhibitors in Small Cell Lung Cancer
PubMed Abstract | CrossRef Full Text | Google Scholar
Graph Neural Networks: A Review of Methods and Applications
CrossRef Full Text | Google Scholar
Zügner, D., and Günnemann, S. (2019). Adversarial Attacks on Graph Neural Networks via Meta Learning. arXiv:1902.08412v1. Available at: https://arxiv.org/abs/1902.08412 (Accessed December 20
Google Scholar
Dias EA and Jose Aguiar Coelho N (2022) Graph Neural Networks as a Potential Tool in Improving Virtual Screening Programs
Received: 18 October 2021; Accepted: 10 December 2021;Published: 20 January 2022
Copyright © 2022 Alves, Ferreira, Maricato, Alberto, Dias and Jose Aguiar Coelho. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use
distribution or reproduction in other forums is permitted
provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited
in accordance with accepted academic practice
distribution or reproduction is permitted which does not comply with these terms
*Correspondence: Luiz Anastacio Alves, YWx2ZXNsYWEzMEBnbWFpbC5jb20=
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations
Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher
94% of researchers rate our articles as excellent or goodLearn more about the work of our research integrity team to safeguard the quality of each article we publish
Please select what you would like included for printing:
Copy the text below and then paste that into your favorite email application
Enter your phone number above to have directions sent via text
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply
Service map data © OpenStreetMap contributors
KOROR (Island Times/Pacnews) — The speaker of the House of Delegates
penalties and interest on real estate transactions he conducted in 2015 for which he failed to obtain a license and pay taxes on real estate income of $1.875 million
penalty and interest assessed against Anastacio is $3.5 million
The court also deferred his prosecution for 24 months or until July 2025
Anastacio at the same time is ordered to obtain a license
obey all laws and pay his taxes by July 2025
has drawn significant public attention and raised questions about transparency and accountability among Palau’s leadership
Anastacio’s legal troubles involved a series of real estate transactions conducted between January 2015 and January 2016
he reportedly earned over $3.27 million from leasing properties in Aimeliik State
One of the key deals was a 49-year lease with Chinese national Qin Zhong that generated a substantial portion of Anastacio’s income
Anastacio failed to obtain a valid business license as required by Palauan law (40 PNC §1501) while conducting these transactions
he is accused of not reporting the income he received from these deals
which mandates that all business-related income be reported for tax purposes
Anastacio faces up to 10 years in prison and fines of up to $25,000 for misconduct in public office
which is classified as a Class B felony under Palauan law
could result in a one-year prison sentence or a fine of $1,000
the court has approved a deferred prosecution agreement
allowing Anastacio to avoid immediate trial
Anastacio must pay the overdue taxes and obtain a valid business license to continue any real estate activities
If he complies with the terms of the agreement and avoids further legal violations
the charges against him could be dismissed in July 2025
This allows the speaker to rectify the situation without immediate criminal prosecution while his case undergoes scrutiny
The allegations against him have sparked concerns about the ethical responsibilities of public officials
including legal experts and watchdog groups
have stressed the need for greater transparency and adherence to legal standards among government leaders
the case highlights the challenges surrounding accountability in Palau’s public sector
as the investigation revealed the complexities involved in enforcing tax laws
particularly when government officials are involved
The Office of the Special Prosecutor has emphasized the importance of holding public servants to the same legal standards as any other citizen
remains open and will be closely monitored over the next two years to ensure that Anastacio complies with the conditions of his deferred prosecution
such as paying the full amount of back taxes or obtaining the necessary business license
the Republic of Palau reserves the right to resume prosecution
potentially leading to the full reinstatement of charges and corresponding penalties
which played a central role in investigating Anastacio’s financial activities
has vowed to remain vigilant in enforcing Palau’s tax laws
The BRT’s findings revealed that Anastacio and his spouse
were engaged in leasing mangrove properties in Aimeliik without proper business partnership or corporation registration
further compounding the legal issues faced by the couple
The case has broader implications for Palau’s leadership and political environment
with some questioning how widespread such practices might be among other public officials
the spotlight remains on whether he will fulfill the legal obligations set out by the court
a matter that many will be watching closely
Anastacio continues in his role as speaker of the House
but his legal entanglements cast a shadow over his tenure
the case serves as a reminder of the critical importance of ethical governance and the need for accountability at the highest levels of Palau’s government
Your browser is out of date and potentially vulnerable to security risks.We recommend switching to one of the following browsers:
Post Courier
Sabino Anastacio was ordered to pay taxes
penalties and interest on real estate transactions he conducted in 2015 where he failed both to obtain a license and pay taxes on real estate income of US$1.875million
penalty and interest assessed against Anastacio is US$3.5 million
Court order deferred prosecution for 24 months or until July 2025
obey all laws and pay off his taxes by July 2025
Anastacio’s legal troubles are rooted in a series of real estate transactions conducted between January 2015 and January 2016
he reportedly earned over US$3.27 million from leasing properties in Aimeliik State
One of the key deals was a 49-year lease with Chinese national Qin Zhong
which generated a substantial portion of Anastacio’s income
Anastacio failed to obtain a valid business license as required by Palauan law (40 PNC §1501)
the Speaker is accused of not reporting the income he received from these deals
including up to 10 years in prison and fines of up to US$25,000 for misconduct in public office
could result in a one-year prison sentence or a fine of US$1,000
the court has reached a deferred prosecution agreement
This type of agreement allows the Speaker a chance to rectify the situation without immediate criminal prosecution but places the case under ongoing scrutiny
Anastacio’s high-profile role as Speaker of the House has amplified public interest in the case
the allegations have sparked concerns about the ethical responsibilities of public officials
The Office of the Special Prosecutor has emphasised the importance of holding public servants to the same legal standards as any other citizen
Anastacio continues in his role as Speaker of the House
With deferred prosecution and taxes unpaid
the case against Anastacio remains active…
Get the latest news delivered straight to your inbox
Send help right to the people and causes you care about
Your donation is protected by the GoFundMe Giving Guarantee
Share on FacebookShare on X (formerly Twitter)Share on PinterestShare on LinkedInWICHITA FALLS
Texas (KAUZ) - Anastacio Mendoza has been indicted in reference to the wreck on Loop 11 that killed 68-year-old Diane Luckett on March 16
Mendoza is being indicted on the charge of Intoxicated Manslaughter with a Vehicle
Medoza has been held in Wichita County Jail since the wreck
Luckett passed away in Lubbock on Saturday from her injuries
Charlie Eipper told our crews that a truck was driving westbound on Seymour Highway while a car turned left from eastbound Seymour Highway onto Loop 11
ran the light at a high speed and hit the car
The woman was in the vehicle turning onto Loop 11
Mendoza was in the truck driving westbound on Seymour Hwy
More information on the wreck can be found here
Inform your decision making with data that supports thousands of decisions daily
Find out how ICIS data is helping businesses in your sector
Meet strategic objectives with specialised analytics that optimise outcomes
Optimise outcomes with expert news and analysis on the issues that matter
Connecting markets to optimise global resources
with unlimited access to ICIS chemicals news across all markets and regions
the industry-leading magazine for the chemicals industry
Partnering with ICIS unlocks a vision of a future you can trust and achieve
We leverage our unrivalled network of industry experts to deliver a comprehensive market view based on independent and reliable data
insight and analytics.Contact us to learn how we can support you as you transact today and plan for tomorrow
ICIS® is part of LexisNexis® Risk Solutions
Copyright © 2025 LexisNexis Risk Solutions
Repped by/in: Cadence Films US, Frenzy Paris
Awards: Best Film at Milan Fashion Film Festival, Philadelphia Museum of Art Official Selection for Design for Different Futures Exhibition, Best Director NYFFF 2018
Barbara> It’s extremely important for me to understand the product, brand, and market I’m creating for. Not only for obvious practical reasons but also on a personal level - I want to feel aligned and involved with what I’m making.
Barbara> I’m passionate about storytelling and the human experience, whatever genre, form, or subject matter that might be. I truly believe everyone, no matter how apparently dull, has a story worth telling. Empathising with another’s plight is at the core of what I do and what truly draws me. Work that dares to achieve this is what excites me.
Barbara> That my work is 'real' and 'authentic'. That’s kind of true to a certain extent but there’s also often a lot of 'staging' that happens before in order to achieve that 'real' look. And I also really enjoy the challenge of creating a fictional world. I do wish I had more opportunities to do so in commercial work.
Barbara> Once while shooting in Oaxaca I had a storm, followed by a mini earthquake, that destroyed our set on the last day of shooting. We had a group of mariachis as part of our cast so we all just took shelter and ended up drinking tequila and singing. It was quite magical and after that, the rainbow came and the bond created by this clearly translated on the footage, even if the set wasn’t as perfect as it should have been.
Barbara> In my experience, the stronger and clearer your vision is, the more respected you are and the less of a balance you’ll need to strike. I think it’s really important to discuss with detail and specificity your vision ahead of time so that by the time you’re on set it’s about executing it together. It’s really important to gain trust and space from the client before in order to bring something to life together
Barbara> That’s not something I hugely focus on. My main focus is to communicate stories in the best way possible. And at the end of the day, making films or videos is about sculpting with time, space, light, and characters. All else follows... I don’t think the formats should dictate the storytelling but vice-versa. A great story will find its way and adapt to multiple forms. I see the different formats as packaging, but the essential part is what’s really inside it.
This website is using a security service to protect itself from online attacks. The action you just performed triggered the security solution. There are several actions that could trigger this block including submitting a certain word or phrase, a SQL command or malformed data.
You can email the site owner to let them know you were blocked. Please include what you were doing when this page came up and the Cloudflare Ray ID found at the bottom of this page.
More fromWork
Contact
Advertising Opportunities
Newsletters
Insights + Opinion
Creatives + Projects
Advice + Resources
Culture + Lifestyle
Nicer Tuesdays
The View From...
POV
Forward Thinking
Review of the Year
Jenny Brewer
Olivia Hingley
Ellis Tree
Elizabeth Goodspeed
Liz Gorny
Extra Search
Barbara is the director with enviable access to some of the finest homes in the creative world
and the project is a calming conversational film between Barbara and the tenant of the home
an anthropology graduate with little experience in film at the time
was approached by the then commissioning director of Nowness Raven Smith to direct the series
but it wasn’t like a definite commission,” she explains
‘We’d like to take you to some people’s houses and see what you come up with,’” explains the director
I really didn’t have a lot of experience at the time.”
In the eighth episode of the series Barbara visits the home of advertising legend George Lois
the first three were just me filming and asking questions and one guy working on the sound,” Barbara tells It’s Nice That
“It was a good way to make people feel confident but it was difficult for me to concentrate on the focus of the camera and be engaging with the person in front of me.” Over time
her team expanded to a director of photography to lend a helping hand
“But still we keep it really small because that is what keeps people open
It’s quite intrusive to go to someone’s house with a camera
Marianne Faithful and Peter Shire to name a few could seem a nerve wracking experience
Barbara doesn’t spend a lot of time digging into their lives beforehand
“For the first ones I really didn’t have time to do so,” she explains
“I’m still not sure what is the best approach
obviously sometimes I think this is a good one to research
This also means that the questions Barbara asks aren’t formally scripted
“There are general ones I ask most people but it’s just reacting to what I see in the space mostly
That and how they feel that day and even how I am feeling
Usually I find it’s more helpful to just feed from that day to see an outcome
The original films were usually apartments in New York
as the series continues larger houses are seen
“There are some that are a little more challenging in the space itself
It’s definitely different to the small apartments I was used to
The elements the director decides to include depends on the space
“In some cases it’s the structure that is interesting
Others it’s the little knick-knacks they have that are interesting.” Barbara’s directorial style is predominately digital in the shorts
but small sections cut to characteristic elements of people’s homes in VHS
This stylistic decision was informed by Barbara’s grandmother’s house
“When the series started I was at my grandmother’s in Brussels and she actually had all these VHS tapes of us
When I was looking through them I thought it was perfect
it brings you back to childhood videos and how you really live in a space
rather than the perfect image of interior design.”
On asking Barbara if there were any favourites during her time creating My Apartamento she answers: “They are all kind of special
It’s really hard not to appreciate everyone.” However one couple
was one Tchaik originally designed for David Hockney
Tchaik and Melissa eventually bought the apartment from David and made it their family home
Barbara’s ability to make the films so personal
as if you’ve actually popped round to this person’s house for lunch
it’s very artisanal to me as a piece,” she explains
“It’s really subtle things that I look for
it’s easier to do it yourself as some editors would cut out those off camera moments
the ones when you see the person off guard and how they really live.”
My Apartamento: George Lois by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Florence Welch by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Kelis by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Peter Shire by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Tchaik Chassay and Melissa North by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Marianne Faithful by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Jean-Charles de Castelbajac by Barbara Anastacio for Nowness and Apartamento
My Apartamento: Adwoa Aboah by Barbara Anastacio for Nowness and Apartamento
Lucy Bourton
Lucy (she/her) was part of the It’s Nice That team from 2016–2025
first joining as a staff writer after graduating from Chelsea College of Art with a degree in Graphic Design Communication
eventually becoming a senior editor on our editorial team
a research-driven department with It’s Nice That
Fancy a bit of It's Nice That in your inbox
Sign up to our newsletters and we'll keep you in the loop with everything good going on in the creative world
Instagram
TikTok
LinkedIn
Facebook
Twitter
Pinterest
About
Careers at It’s Nice That
Privacy Policy
Insights
Residence
Creative Lives in Progress
If You Could Jobs
© It’s Nice That 2024 · Nice Face Logo © It’s Nice That
www.barbaranastacio.com\nwww.nowness.com\nwww.apartamentomagazine.com
Barbara Anastacio is a name you may recognise in the opening credits to My Apartamento, an interiors series from Nowness
this custom title design balances broken structures with ornate details to embody the protagonist’s sense of disconnection
We headed to Arsenal’s training ground to chat to the defender about using creativity to forge a bridge between footballers and the fans
Porous is a beautifully tender and tactile meditation on healing from sexual trauma
Following the release of his debut book, New York Nico’s Guide to NYC
the director and documentary filmmaker Nicolas Heller sat down with our US editor at large Elizabeth Goodspeed to discuss the origins of his widely loved Instagram page
his career in filmmaking thus far and why he’s intent on capturing the city’s most charismatic characters
About
Contact
Advertising Opportunities
Newsletters
Insights + Opinion
Creatives + Projects
Advice + Resources
Culture + Lifestyle
Nicer Tuesdays
The View From...
POV
Forward Thinking
Review of the Year
Jenny Brewer
Olivia Hingley
Ellis Tree
Elizabeth Goodspeed
Liz Gorny
Instagram
TikTok
LinkedIn
Facebook
Twitter
Pinterest
Careers at It’s Nice That
Privacy Policy
Insights
Residence
Creative Lives in Progress
If You Could Jobs
2021 at 2:45 pm CT.css-79elbk{position:relative;}Anastacio Salazar
of Joliet was arrested by Joliet police late Wednesday night following a domestic violence outburst against his future wife
(Mugshot via Joliet Police Department )JOLIET
IL —A 28-year-old Joliet man attacked a woman
forced her head into a wall and kitchen table in the 300 block of Wheeler Avenue on Wednesday night
The woman got away and ran out of the house
flagging down a Joliet police officer patrolling the neighborhood
Anastacio Salazar was charged with aggravated domestic battery
interfering with the reporting of domestic violence
an officer was flagged down shortly before 11 p.m
near West Marion and Oneill Streets and "officers determined that a female victim had been battered by her fiancé
at a residence in the 300 block of Wheeler Avenue."
Police said "Salazar attacked the victim multiple times
choking her and forcing her head into a wall and kitchen table
Salazar refused to let the victim leave the residence."
Salazar grabbed the woman's phone throwing it to the ground
When the woman ran from the house and tried to drive away
"Salazar exited the residence and began punching her vehicle and jumping on the hood
The victim was able to flee Salazar and flag down an officer on patrol in the area," police reports state
The woman refused medical treatment for her injuries
who was arrested by Joliet police in the same area
"Salazar was then transported to Amita St Joseph Medical Center by the Joliet Fire Department after complaining about an injury sustained during the incident," police reports show
"Salazar was transported to the Will County Jail upon release from the hospital."
who lives in Joliet in the 200 block of Westport Drive
was booked into the Will County Jail around 3:20 a.m
Get more local news delivered straight to your inbox. Sign up for free Patch newsletters and alerts.
Please include what you were doing when this page came up and the Cloudflare Ray ID found at the bottom of this page.