Bioinformatics and Biomedical Engineering

Protein structures are complex and dynamic entities relevant to many biological processes. G-protein-coupled receptors in particular are a functionally relevant family of cell membrane proteins of interest as targets in pharmacology. Nevertheless, the limited knowledge about their inherent dynamics hampers the understanding of the underlying functional mechanisms that could benefit rational drug design. The use of molecular dynamics simulations and their analysis using Machine Learning methods may assist the discovery of diverse molecular processes that would be otherwise beyond our reach. The current study builds on previous work aimed at uncovering relevant motifs (groups of residues) in the activation pathway of the $$\beta 2$$ β 2 -adrenergic ( $$\beta _2AR$$ β 2 A R ) receptor from molecular dynamics simulations, which was addressed as a multi-class classification problem using Deep Learning methods to discriminate active, intermediate, and inactive conformations. For this problem, the interpretability of the results is particularly relevant. Unfortunately, the vast amount of intermediate transformations, in contrast to the number of re-orderings establishing active and inactive conditions, handicaps the identification of relevant residues related to a conformational state as it generates a class-imbalance problem. The current study aims to investigate existing Deep Learning techniques for addressing such problem that negatively influences the results of the predictions, aiming to unveil a trustworthy interpretation of the information revealed by the models about the receptor functional mechanics.

Drug development is a complex process that remains subject to risks and uncertainties. In its early days, much emphasis was placed on the equilibrium binding affinity of a drug to a particular target, which is described by the equilibrium dissociation constant ( $$K_{d}$$ K d ). However, there are a large number of drugs that exhibit non-equilibrium binding properties. For this reason, optimization of other kinetic parameters such as dissociation constants ( $$k_{off}$$ k off ) and association constants ( $$k_{on}$$ k on ) is becoming increasingly important to improve accuracy in measuring in vivo efficacy. To achieve this, the concept of residence time between drug and target ( $$\tau $$ τ ) was developed to account for the continuous elimination of the drug, the absence of equilibrium conditions, and the conformational dynamics of the target molecules. Residence time has been shown to be a better estimate of drug lifetime potency than equilibrium binding affinity and is recognized as a key parameter in drug development. However, because residence time is only one measure of drug potency, it provides only a limited picture of binding kinetics and affinity.A machine-learning algorithm was proposed to identify molecular features affecting protein-ligand binding kinetics for a set of similar compounds. Molecular dynamics simulations of $$\tau $$ τ RAMD results were used as model input. The study confirmed that $$\tau $$ τ RAMD provides information about the characteristics of the dissociation pathway since the obtained dissociation trajectories can be used to identify the interactions that occur and the conformational changes of the system at subsequent time points. The proposed algorithm made it possible to obtain information on protein-ligand contacts that are specific to their residence times.

The WHO's standards for analysis and presentation of anthropometric data is widely recognized as the best system. This study was aimed to assess the agreement of calculation of height-for-age and BMI-for-age Z-scores according to WHO growth charts using of two randomly selected mobile apps in two cohorts of European and Asian subjects. In 1,347 adolescents aged 13 to 17 years, boys and girls, living in St. Petersburg, Northwest of the Russia (European cohort, 663 subjects) and Nukus, Uzbekistan (Asian cohort 684 subjects) measured body weight and stature. Each child's height-for-age Z-score and BMI-for-age Z-scores were calculated based on WHO Child Growth Standards PC version WHO AntroPlus software and mobile applications. It was performed a Blend-Altman analysis to assess the consistency of Z-scores calculated using WHO AntroPlus software and two mobile apps. The results of the evaluation of the consistency of height-for-age and BMI-for-age Z-scores obtained using WHO AntroPlus software and mobile applications showed that Android-based mobile applications systematically overestimate Z-scores. However, this error is much smaller than any clinically significant deviations of the physical development evaluation parameters both in the European and Asian cohort volunteers. Thus, it can be claimed that it is possible to use Android-based mobile applications for growth monitoring and BMI and WHO AntroPlus software for children from 5 to 19. This is especially relevant to the need for «field» monitoring of children's growth and development performed by general practitioners visiting patients at home.

A challenge in data analysis in bioinformatics is to offer integrated and modern access to the progressively increasing volume of data, as well as efficient algorithms for their processing. Considering the vast databases of biological data available, it is extremely important to develop efficient methods for processing biological data. A new algorithm for arranging DNA sequences based on the suggested CAT method is proposed, consisting of an algorithm for calculating a CAT profile against the selected reference sequences and an algorithm for comparing two sequences, based on the calculated CAT profiles. Implementation steps, inputs and outputs are defined. A software implementation of the proposed method for arranging biological sequences CAT has been designed and developed. Experiments have been carried out using different data sets to align DNA sequences based on CAT method. An analysis of the experimental results have been done in terms of collisions, speed and effectiveness of the proposed solutions.

Computing maximal perfect blocks of a given panel of haplotypes is a crucial task for efficiently solving problems such as polyploid haplotype reconstruction and finding identical-by-descent segments shared among individuals of a population. Unfortunately, the presence of missing data in the haplotype panel limits the usefulness of the notion of perfect blocks.We propose a novel algorithm for computing maximal blocks in a panel with missing data (represented as wildcards). The algorithm is based on the Positional Burrows-Wheeler Transform (PBWT) and has been implemented in the tool Wild-pBWT, available at https://github.com/AlgoLab/Wild-pBWT/ . Experimental comparison showed that Wild-pBWT is 10–15 times faster than another state-of-the-art approach, while using a negligible amount of memory.

The world of computing is constantly evolving. The trends that are shaping today’s applications are Cloud computing and GPU computing. These technologies allow bringing high performance computations to low power devices, when using a computing outsourcing architecture. Following the trend, bioinformatic applications are looking to take advantage of these paradigms, but there are challenges that have to be solved. Data that these applications work with is usually sensible and has to be protected. Also, GPU usage in Cloud architectures currently presents inefficiencies. This paper makes a review of the characteristics of Cloud computing outsourcing architectures, including the security aspects, and GPU usage for these applications. The proposed architecture includes GPU devices and tries to make efficient use of them. The experiments show that it has the opportunity to increase parallelism and reduce context switching costs when running different applications concurrently on the GPU.

This article presents a study that aimed to identify the stages of bladder cancer based on gene expression data. The dataset used in the study was obtained from the GDC repository and included 406 cases of bladder cancer and 431 files from the TCGA-BLCA project. The study categorized the cases into three classes based on disease stages: Stage 2, Stage 3, and Stage 4. The methodology employed R programming language and the KnowSeq library for the study development. The authors identified genes that showed significant differences in expression among the classes and created a matrix of differentially expressed genes (DEG). Machine learning models, including feature selection algorithms and classification models such as KNN and SVM, were constructed to predict the bladder cancer stages. The results revealed that the mRMR feature selection algorithm performed the best, and the 8 most relevant genes were used to build the classification models.

In recent years, serum-based tests for early detection and detection of tissue of origin are being developed. Circulating microRNA has been shown to be a potential source of diagnostic information that can be collected non-invasively. In this study, we investigate circulating microRNAs as predictors of cancer stage. Specifically, we predict whether a sample stems from a patient with early stage (0-II) or late stage cancer (III-IV). We trained five machine learning algorithms on a data set of cancers from twelve different primary sites. The results showed that cancer stage can be predicted from circulating microRNA with a sensitivity of 71.73%, specificity of 79.97%, as well as positive and negative predictive value of 54.81% and 89.29%, respectively. Furthermore, we compared the best pan-cancer model with models specialized on individual cancers and found no statistically significant difference. Finally, in the best performing pan-cancer model 185 microRNAs were significant. Comparing the five most relevant circulating microRNAs in the best performing model with the current literature showed some known associations to various cancers. In conclusion, the study showed the potential of circulating microRNA and machine learning algorithms to predict cancer stage and thus suggests that further research into its potential as a non-invasive clinical test is warranted.

Gait analysis plays a crucial role in medical diagnostics due to its ability to determine and quantify the patient’s physical abilities and limitations. Unlike the other competing sensors, radar is capable of measuring human gait in non-contact fashion. In this paper, we present the extraction of six different gait parameters using Frequency Modulated Continuous Wave (FMCW) radar within five walking steps. The range-time and Doppler-time information are used to extract the parameters. The range-time information of FMCW radar yields the walking duration, walking time, and average walking velocity whereas, the velocity-time information yields pause time during walking, inter-step distance variation, and inter-step time variations. An Inertial Measurement Unit (IMU) is deployed as a ground-truth reference sensor to track the gait movement and a high correlation is found between radar and the reference sensor. Finally, as a use case example, gait parameters analysis is performed to detect asymmetric gait movement. Symmetric and asymmetric walking data is collected with radar and features analysis is performed which suggests that inter-step time and velocity variations contributes greatly in asymmetry detection.

Magnetic resonance imaging (MRI) technology is rapidly advancing and three-dimensional (3D) scanners started to play an important role on diagnosis. However, not every medical center has access to 3D magnetic resonance imaging (MRI) devices; therefore, it is safe to state that the majority of MRI scans are still two-dimensional. According to the setup values adjusted before any scan, there might be consistent gaps between the MRI slices, especially when the increment value exceeds the thickness. The gap causes miscalculation of the lesion volumes and misjudgments when the lesions are reconstructed in three-dimensional space due to excessive interpolation. Therefore, in this paper, we present the details of three types of conventional morphing methods, one dilation-based and two erosion-based, and compare them to figure out which one provides better solution for filling up the gaps in incremental brain MRI. Among three types of morphing methods, the highest average dice score coefficient (DSC) is calculated as %91.95, which is obtained by the multiplicative dilation morphing method for HG/0004 set of BraTS 2012.

Three-dimensional (3D) visualization of the brain tumors reconstructed from the two-dimensional (2D) magnetic resonance imaging (MRI) sequences plays an important role in volumetric calculations. The reconstructions are usually executed using the fluid attenuated inversion recovery (FLAIR) sequences, where the whole tumors appear brighter than the healthy surrounding tissues. Without any processing; however, reconstruction results might be inconclusive; therefore, we propose a mathematical m-parametric Nakagami imaging for highlighting the lesions. The raw 2D FLAIR MRI images are taken from BraTS 2012 dataset and the highlighted images are generated by the Nakagami imaging. The information on the MRI slices is compiled in three-layered Nakagami images for better visualization of the high-grade and low-grade glioma in 3D space. By the flexible m-parametric design, on the other hand, the reconstructed images might easily be adjusted according to the GT images for precise representation.

Radiotherapy is one of the most widely used treatments for cancer by irradiating the tumor volume. However, one of its disadvantages is that healthy tissue is also affected, producing various side effects. For this reason, preliminary studies are required beforehand to determine the dose to be administered in each case, to avoid possible damage and to make sure that the dose received by the tumor is the correct one. These studies are carried out both using simulations and with routine machinery procedures using a mannequin that simulates the area to be treated. In this work a way of speeding up the previous study process is tackled, starting from simulated data whose optimized obtaining will be the objective of this work. The PENELOPE Monte Carlo simulation software is used to recreate the process and obtain the necessary previous data. Subsequently, regression models are applied to obtain the values of interest and accelerate the procedure, reducing, in addition, the energy consumption and storage required while obtaining very accurate approximations.

The COVID-19 pandemic highlighted the importance of RNA-based technologies for the development of new vaccines. Besides vaccines, a world of RNA-based drugs, including small non-coding RNA, could open new avenues for the development of novel therapies covering the full spectrum of the main human diseases. In the context of the “National Center for Gene Therapy and Drugs based on RNA Technology” funded by the Italian PNRR and the NextGenerationEU program, our lab will contribute to the construction of a Knowledge Graph (KG) for RNA-drug analysis and the development of innovative algorithms to support RNA-drug discovery. In this paper, we describe the initial steps for the identification of public data sources from which information about different kinds of non-coding RNA sequences (and their relationships with other molecules) can be collected and used for feeding the KG. An in-depth analysis of the characteristics of these sources is provided, along with a meta-graph we developed to guide the RNA-KG construction by exploiting and integrating biomedical ontologies and relevant data from public databases.

Proteins are active players in different sarcomas through actively participating in downstream signaling pathways. EWS RNA-Binding Protein 1 (EWSR1) is considered as good therapeutic target for treatment of Ewing’s sarcoma. In current study, comparative modeling approach was employed to predict the structural models of three different domains of EWSR1 and Domain Enhanced MOdeling (DEMO) server was used to compose the predicted models of three domains. Furthermore, RNA motifs binding to EWSR1 were predicted and the 3D model of RNA bound to EWSR1 was built by using MC-Fold. The conformational interactions between EWSR1 and RNA were studied with HNADOCK. Moreover, MD simulations were performed to check the stability of EWSR1-RNA complex by computing RMSD, RMSF, Rg, and SASA graphs. Based on computational assessments, it has been concluded that certain structural and dynamical features of EWSR1-RNA complex could be used as a target in future development of drugs against Ewing’s sarcoma.

Drug design is highly priced and time-consuming procedure. Therefore, to overcome this problem, in silico drug design methods, particularly drug repositioning have been developed to assess therapeutic effects of known drugs against other diseases. In this study, computational drug repositioning method is used to explore the alternative therapeutic effects of FDA approved drugs to treat Alzheimer’s disease. The chemical shape-based screening was employed to fetch some potential new drugs based on the structure of standard drugs. The screened drugs were further evaluated through pharmacogenomics, molecular docking, and molecular dynamics simulation studies. The best lead drugs, such as darifenacin, astemizole, tubocurarine, elacridar, sertindole and tariquidar displayed promising repositioned effects in the treatment of Alzheimer’s disease and may be used as potential medicines after thorough experimental and clinical studies.

The study of the movement of the vertebrae of the lumbar spine is classified as a relevant theme for research, considering the possibility of exploring the pathological dysfunctions of this region. This paper presents the development of a trajectory motion control for a lumbar spine model. The spine model is being represented by a 2 DOF (Degrees of Freedom) manipulator robot, which represents the motion of two lumbar vertebrae. For the computational simulations of the controlled spine behavior the mathematical dynamic model of the manipulator based on the Lagrange approach is being considered. Preliminary simulation results show that the implemented conventional controller robustly follows the references given for the angles of the vertebrae, guaranteeing the planned movement.

Orthology assignment between genetic elements lies at the heart of comparative genomics. Current methods primarily rely on sequence and structural similarity. Both low sequence similarity and the presence of multiple copies limit similarity-based methods. Synteny, i.e., conservation of (relative) genomic location, can help resolve many such cases. The mapping of synteny is based on “synteny anchors”, defined as intervals of genomic locations for which unique orthologs in related genomes can be determined unambiguously. Usually, annotated elements such as protein-coding genes are utilized for this purpose. Here we describe an annotation-free approach and devise a k-mer-based heuristic to identify synteny anchors. To demonstrate the practicality of the approach, we compute and analyze a set of synteny anchors for 25 Drosophila species.

Breast Cancer is no longer considered a death sentence and the number of patients receiving radiotherapy is increasing every year. Advancements in RT technology allow more accurate and accumulated delivery of radiation significantly reducing the risk of side effects. The use of these techniques has led to better clinical outcomes and significantly increased long-term survival rates. Therefore, secondary cancer risk after breast conserving therapy is becoming more important. In this study, we estimate the risks of developing a solid second cancer after radiotherapy of breast cancer using the concept of organ equivalent dose (OED).Eight breast cancer patients were retrospectively selected for this study. Three-dimensional conformal radiotherapy (3DCRT), intensity modulated radiotherapy (IMRT), and volumetric modulated arc therapy (VMAT) were planned to deliver a prescribed dose. Differential dose volume histograms (dDVHs) were created and the OEDs calculated. Secondary cancer risks of ipsilateral, contralateral lung and thyroid gland were estimated using linear, linear-exponential and plateau models.The highest interest of our study was the evaluation of secondary cancer risk for the organs at risk (OAR), which are located far from the treatment region and are very sensitive to radiation exposure. Our results showed very high secondary cancer excess absolute risk (EAR) values for IMRT and VMAT compared with 3DCRT. It has to be noted, that a significant reduction of the EARs for the contralateral lung, ipsilateral lung and thyroid gland was observed in all dose–response models.

The objective of this work is to present a radiomics-based platform (RAB-PET) and method to detect Alzheimer’s disease (AD) non-invasively using 18FDG-PET images. Radiomic analysis allows the identification of regional features that serve to predict the presence or characteristics of diseases using images as data. We first, used the FastSurfer a deep learning-based toolbox to segment the whole brain into 95 classes by the utilization of the DKT-atlas. Then the PyRadiomics toolbox was used to extract features from 18FDG-PET scans. After preprocessing, the features were subject to a selection process by making use of eight different methods, namely, ANOVA, PCA, Chi-square, LASSO, Recursive Feature Elimination (RFE), Feature Importance (FI), Mutual Information (MI), and Recursive Feature Addition (RFA). Finally, in order to classify the selected features by feature selection methods, we implemented nine different classifier methods, namely, GradientBoosting (GB), RandomForest (RF), DecisionTree (DT), GaussianNB (GNB), GaussianProcess (GP), MLP, QuadraticDiscriminantAnalysis (QDA), AdaBoost (AB), and KNeighbors (KNN) on selected feature subsets. All data (scans and clinical examination results) were obtained from the AD Neuroimaging Initiative (ADNI) database. The RF classifier with 100 iterations on features obtained with the LASSO algorithm yielded an area under the curve of AUC = 0.976 with a 95% confidence interval of 0.93–0.98 based on 30% independent test data. We conclude that a platform for radiomic analysis can serve as a potential method for deducing accurate information on brain diseases such as Alzheimer’s disease non-invasively using 18FDG-PET images. Further studies are underway to extend this work by studying the association between the set of features and several characteristics of the Alzheimer’s disease.

Automated electroencephalographic (EEG) signals classification using deep learning algorithms is an emerging technique in neuroscience that has the potential to detect brain pathologies such as epilepsy efficiently. In this process, deep learning algorithms are trained with labeled EEG signal datasets. However, due to the highly complex nature of EEG signals and the large amount of irrelevant information they contain, feature extraction techniques must be applied to reduce their dimensionality and focus on relevant information. This paper presents a comparative study on feature extraction methods for the classification of EEG recordings. The results demonstrate that the proposed classification algorithms and characterisation techniques are effective and suitable, as the accuracy metrics reach a value of 99.27%. The results presented in this paper contribute to the further development of automatic EEG signal classification methods based on deep learning.

Drug stability describes its ability to maintain physical, chemical, therapeutic, and microbiological properties during storage. The study aimed to assess whether there are any differences in the homogeneity of the effervescent tablets containing quercetin and calcium using hyperspectral imaging and X-ray microtomography. We analyzed unexpired, expired, as well as stressed tablets which were stored at 40 ℃ for 14 days. Both unexpired and expired tablets met the pharmacopoeial requirements. The homogeneity analysis showed significant differences between the three types of tablets. In addition, significantly higher reflectance values in the visible and near-infrared range were observed in unexpired tablets compared to expired and stressed tablets (p = 0.001 and p < 0.001, respectively). The X-ray microtomography showed that the densities of unexpired and stressed tablets were significantly higher compared to the density of expired tablets (p < 0.001). The changes in tablets' homogeneity may indicate possible physical changes in effervescent tablets during storage, especially in conditions deviating from those recommended by the manufacturer.

The aim of this study was to investigate the impact of biofeedback (BFB) on manual wheelchair learning. The researchers conducted training sessions with two groups of participants, one using BFB and the other group without it (NBFB). The hypothesis was that BFB would reduce the learning time and help participants to achieve balance positions more quickly. The study enrolled 24 participants aged 24 ± 6 years old; they were divided into two groups of 12 subjects each (BFB and NBFB). The researchers also collected additional information about the participants, such as the sport they practiced, for future investigations. The data was collected using a non-contact electronic angular system placed directly on the wheelchair, measuring spatiotemporal parameters such as the angle between the wheelchair and the ground and the time at which this angle is reached. The results which are statistically significant (p < 0.05) were only obtained between early falling, learning time and number of trials. The study found that BFB did not seem to accelerate the learning time for the wheelie skill on manual wheelchair (BFB group). However, the BFB method could potentially reduce the number of trials using the manual wheelchair under (NBFB). In conclusion, the study showed that biofeedback may not necessarily accelerate the learning time for the wheelie skill on manual wheelchair but can help individuals to maintain balance positions with fewer trials. Further studies are required to confirm these results, as they only involved a small sample size. This study highlights the potential for using biofeedback as an effective tool for wheelchair training and could improve the quality of life of individuals with mobility impairments.

An electrocardiogram (ECG) is a simple test that checks the heart’s rhythm and electrical activity and can be used by specialists to detect anomalies that could be linked to diseases. This paper intends to describe the results of several artificial intelligence methods created to automate identifying and classifying potential cardiovascular diseases through electrocardiogram signals. The ECG data utilized was collected from a total of 46 individuals (24 females, aged 26 to 90, and 22 males, aged 19 to 88) using a BITalino (r)evolution device and the OpenSignals (r)evolution software. Each ECG recording contains around 60 s, where, during 30 s, the individuals were in a standing position and seated down during the remaining 30 s. The best performance in identifying cardiovascular diseases with ECG data was achieved with the Naive Bays classifier, reporting an accuracy of 81.36%, a precision of 26.48%, a recall of 28.16%, and an F1-Score of 27.29%.

Discovering functionalities for unknown enzymes has been one of the most common bioinformatics tasks. Functional annotation methods based on phylogenetic properties have been the gold standard in every genome annotation process. However, these methods only succeed if the minimum requirements for expressing similarity or homology are met. Alternatively, machine learning and deep learning methods have proven helpful in this problem, developing functional classification systems in various bioinformatics tasks. Nevertheless, there needs to be a clear strategy for elaborating predictive models and how amino acid sequences should be represented. In this work, we address the problem of functional classification of enzyme sequences (EC number) via machine learning methods, exploring various alternatives for training predictive models and numerical representation methods. The results show that the best performances are achieved by applying representations based on pre-trained models. However, there needs to be a clear strategy to train models. Therefore, when exploring several alternatives, it is observed that the methods based on CNN architectures proposed in this work present a more outstanding facility for learning and pattern extraction in complex systems, achieving performances above 97% and with error rates lower than 0.05 of binary cross entropy. Finally, we discuss the strategies explored and analyze future work to develop integrated methods for functional classification and the discovery of new enzymes to support current bioinformatics tools.

Bioinformatics is the application of computer science and information technology to the field of biology and medicine. It involves the analysis of large amounts of biological data, such as DNA sequences, protein structures, and gene expression patterns. Bioinformatics is used to develop new methods for understanding and analyzing biological data, as well as to develop new tools and technologies for biological research. Bioinformatics is used in a variety of fields, including genomics, proteomics, and drug discovery. In this study, focus on two severe diseases which affect millions of people globally such as stress and type 2 diabetes. Stress can have a significant impact on people with type 2 diabetes. Stress can cause blood sugar levels to rise, making it difficult to manage diabetes. The purpose of this research is to use various bioinformatics methods to discover potential therapeutic drugs and functional pathways between stress and type 2 diabetes. The microarray datasets GSE183648 and GSE20966 are used for the analysis of stress and type 2 diabetes samples respectively. After the datasets have been preprocessed and filtered through the use of the R programming language, identified the common DEGs. The depiction of common DEGs is shown by venn diagram. Next, the most active genes are identified through topological properties, and PPIs are built from the similar differential expressed genes (DEGs). These five genes NTRK2, SOCS3, NEDD9, MAP3K8, and SIRPA are the most important hub genes with in the interaction network of protein-protein. According to the common DEGs, GO terms molecular function (MF), KEGG and WikiPathways are shown in this study. Gene-miRNA interaction, TF-gene regulatory network, module analysis, GO terms (Biological Process, Cellular Component), Pathways (Reactome, BioCarta, BioPlanet) are all things that could be done with this research work in the future. In last, a therapeutic drug compounds are recommended on the basis of common DEGs.

Tyrosine kinases are enzymes that phosphorylate tyrosine residues in specific substrates, and their activities are involved in the pathophysiology of cancer. The inhibitors of tyrosine kinases block their oncogenic activation in cancer cells, therefore presenting a target for the development of new anticancer drugs. The computational methods in drug discovery and development minimize the time and cost needed in drug designing process. We have reviewed the recent advance in the quantitative structure-activity relationship (QSAR) study and molecular docking related to the new antitumor agents, such as amidine derivatives of 3,4-ethylenedioxythiophene, quinoline-arylamidine hybrids, 7-chloro-4-aminoquinoline-benzimidazole hybrids, rhodanine derivatives, and flavonoids isolated from the leaves of Cupressus sempervirens. The QSAR studies revealed important physicochemical and structural requirements for the antitumor activity and generated models for the prediction of antitumor activity of future potent molecules. Molecular docking allows rapid screening of a large number of compounds to determinate of potential binders of the target protein or enzyme, which is related to the anticancer activity and possible mechanism of action.

We introduce the Coherent Multi-representation Problem (CMP), whose solutions allow us to observe simultaneously different geometrical representations for the vertices of a given simple graph. The idea of graph multi-representation extends the common concept of graph embedding, where every vertex can be embedded in a domain that is unique for each of them. In the CMP, the same vertex can instead be represented in multiple ways, and the main aim is to find a general multi-representation where all the involved variables are “coherent” with one another. We prove that the CMP extends a geometrical problem known in the literature as the distance geometry problem, and we show a preliminary computational experiment on a protein-like instance, which is performed with a new Java implementation specifically conceived for graph multi-representations.

Intrinsically Disordered Regions (IDRs) even though they cannot form a defined three-dimensional structure play a pivotal role in modulating cellular processes and signalling pathways. In the present study, we analyse the conformational changes in IDRs upon complex formation using a non-redundant dataset of binary, X-ray solved 356 protein-protein (P-P) complexes and their corresponding unbound forms. IDRs are prevalent in both unbound and complex proteins and after comparing them in both groups they were categorised into three classes: (a) Disordered-Ordered (D-O), where IDRs present in first group were observed to be ordered in the second group (b) Disordered-Partial Ordered (D-PO), where IDRs present in the first group were found to be partially ordered in the second group and (c) Disordered-Disordered (D-D), where IDRs present in one group remained disordered in the other group. The study of secondary structures of residues in the D-O category reveals that majority of IDRs upon complexation form coils followed by helices and strands. Though majority of residues of IDRs in the D-O class are located at the surface of P-P complexes, we observe a significant number of residues form the interface suggesting that they contribute to the stability of the complexes. Amino acids of IDRs under the D-O category are also involved in polar interactions making hydrogen bonds with other residues as well as water. There are some structured and partially structured regions in the unbound proteins which upon complexation become completely disordered. These findings provide fundamental insights into the underlying principles of molecular recognition by disordered regions in P-P complexes.

Machine learning creates new opportunities for medicine and public health, especially in the field of helping medical examiners. The prospect of generating hints for the diagnosis of a particular disease for physicians is widely considered in this field. The work presents only the possibility of predicting coronary heart disease (CHD), which is classified as a civilization disease, that threatens an increasing number of people in the world. The dataset consisting of more than 11,000 patient records was used for the study. In total each record of the dataset contained 16 features (which were analysed by the model) such as BMI, weight, total cholesterol, years of smoking. Several families of regressors were browsed in search of the best fit. Starting from different variants of the linear model, with various regularisation factors, to support vector machines, ensemble learning (RandomForest) and tree-based solutions with gradient boosting. It turned out that extreme gradient boosted trees (XGBoost) proved to provide the best results. After the model was trained a series of backward searches, such as a visual analysis of the tree models, model estimated weights and an additional LIME-based (Local Interpretable Model-agnostic Explanations) explanation were performed.The obtained results look promising in terms of a deployment of this model, but also similar models aimed in different diseases, in software supporting public health professionals. Such machine learning solution can serve as an as an assistant for a physician, which can be a part of a bigger medical data management system.

Comprehensive analysis of lipids is becoming a forefront of clinical data analysis. Due to significant technical advancements, lipidomics is emerging in clinical diagnostics for improvement and earlier detection of a broad range of diseases. However, in order to understand the biological complexities and interrelationships between the molecules, it is important to have a correct representation of the data and visualizations that enable good interpretability of the lipidomic data. Therefore, the present study systematically compares different visualization methods for lipidomic data, based on different computational relations between the selected lipids and supplemented with known biological information. Networks were reconstructed, and an analysis was performed to objectively compare the visualizations.

This paper aims to explore the use of different deep learning techniques, specifically convolutional neural networks (CNNs), for dietary assessment through image food recognition and compare their performance to the human visual system (HVS). Currently, there are three main techniques for using CNNs in this task: training a network from scratch; using an off-the-shelf pre-trained network; and performing unsupervised pre-training with supervised adjustments. In this study, the authors evaluate the performance of three CNN models with varying numbers of parameters (5,000 to 160 million) based on dataset size and spatial image context.The authors also consider human knowledge and classification to compare the performance of the CNNs to the HVS. They find that while the CNNs make errors across different food classes, the HVS tends to make semantic errors with specific food classes. As a result, the HVS shows more consistency in its answers. Overall, the findings suggest that the HVS is more accurate when the dataset is diverse, while the CNN performs better when the dataset is focused on a particular niche.In conclusion, this study provides empirical evidence that machine learning can be more efficient than the HVS in certain tasks but also highlights the strengths and limitations of both approaches. The authors suggest that combining CNNs with other classification techniques, such as bag-of-words, may be a promising approach for improving the accuracy of dietary assessment through image food recognition.

This article reports on the development of a Web platform for the study of Adverse Drug Events (ADEs). The platform is able to import ADE episodes from official Web sites, like OpenFDA, analyse the chemistry of the drugs involved, together with patient data, and produce a potential explanation based on the drugs interactions. Each study uses chemical knowledge to enrich the information on the molecules involved in the episodes. Data Mining is then used to construct models that can help in the explanation of the ADE occurrence and to predict future events. This paper reports on the Web portal developed and the Data Mining experiments conducted to evaluate the quality, and potential explanations of the forecasted adverse reactions, using real reports of drug administration and the subsequent adverse events. The results showed that it was possible to predict the outcomes of ADEs based on the structure of the molecules of the drugs involved and the data collected from real reports of drug administration up to an accuracy of 79%, while also predicting, with high accuracy, the severity of events where the outcome is the death of the patient (with a precision of 98.9%). The platform provides a less expensive and more accurate way of predicting adverse drug reactions compared to traditional methods. This study highlights the importance of understanding drug interactions at a molecular level and the usefulness of utilising Data Mining techniques in predicting ADEs.

Preterm infants are prone to several neurodevelopmental impairments (NDI). Early and accurate diagnosis could cooperate in the treatment of their clinical manifestations. Clinical data from a cohort of preterm infants at the neonatal intensive care unit (NICU) of the Hospital Puerta del Mar, Cadiz, Spain, was used in this work to perform a classification task to predict abnormal magnetic resonance imaging (MRI) findings using machine learning models. The results in this analysis indicate that the best model able to predict abnormal MRI findings was the K-nearest Neighbor (KNN), with a recall of 0.80. This study represents an initial step towards developing a practical and reliable tool for predicting abnormal MRI findings in preterm infants using readily available clinical data.

A healthy brain clears different types of debris with the help of specialized glial cells. These cells contiguously tile the entire central nervous system (CNS), exert many essential complex functions in the healthy CNS, and maintain a healthy balance in the brain. However, over age, these cells fail to control the healthy balance of the proteins and cause different neurodegenerative diseases, one of which is Alzheimer’s disease (AD). In AD, insoluble amyloid-beta plaques accumulate in the extracellular space along with neurofibrillary tangles (NFTs) inside the brain cells. In this paper, we have developed a model and studied the accumulation of amyloid-beta plaques and NFTs along with an anti-amyloid-beta therapy applied in the treatment of the disease. Based on these studies, we have demonstrated the dynamics of the modelling therapy such that the drug helps clear a subsequent amount of amyloid-beta plaques in each dose. Numerical simulations have been used to show different long-term outcomes of the model. To further analyze the disease progression in the brain and its treatment, we have integrated brain connectome data in the network model as part of our developed modelling framework.

The cardiac operating room (OR) is a high-risk, high-stakes environment inserted into a complex socio-technical healthcare system. During cardiopulmonary bypass (CPB), the most critical phase of cardiac surgery, the perfusionist has a crucial role within the interprofessional OR team, being responsible for optimizing patient perfusion while coordinating other tasks with the surgeon, anesthesiologist, and nurses. The aim of this study was to investigate objective digital biomarkers of perfusionists’ workload and stress derived from heart rate variability (HRV) metrics captured via a wearable physiological sensor in a real cardiac OR. We explored the relationships between several HRV parameters and validated self-report measures of surgical task workload (SURG-TLX) and acute stress (STAI-SF), as well as surgical processes and outcome measures. We found that the frequency-domain HRV parameter HF relative power – FFT (%) presented the strongest association with task workload (correlation coefficient: −0.491, p-value: 0.003). We also found that the time-domain HRV parameter RMSSD (ms) presented the strongest correlation with perfusionists’ acute stress (correlation coefficient: −0.489, p-value: 0.005). A few workload and stress biomarkers were also associated with bypass time and patient length of stay in the hospital. The findings from this study will inform future research regarding which HRV-based biomarkers are best suited for the development of cognitive support systems capable of monitoring surgical workload and stress in real time.

The neonatal period is a critical stage where physiological adaptations for extra-uterine life occur, and newborns are vulnerable to various diseases and disorders. Among these conditions, preterm neonates (PN) born before 37 weeks’ gestation are at higher risk of developing intraventricular hemorrhage (IVH), a common complication that can result in severe neurological complications such as cerebral palsy, developmental delays, and cognitive impairments. Early detection and intervention are essential to prevent long-term consequences.Non-invasive cardiac output monitors (NICOM) have been widely accepted in the neonatal intensive care unit (NICU) for monitoring hemodynamic parameters and have provided vast amounts of data. However, further research is required to explore their predictive tendencies in relation to IVH.The present study aimed to evaluate the potential of deep learning models to enhance early detection and prevention of IVH in preterm neonates using NICOM parameters. From this study, it was shown that by the LSTM autoencoders are able to predict IVH with moderate precision and accuracy but poor specificity. Nonetheless, this study represents a significant step towards developing a non-invasive, accurate, and timely method for monitoring and preventing IVH in preterm neonates, especially in low-resource settings.

Acute pain is a frequent symptom in children who access the Emergency Department (ED). Its measurement through validated tools compatible with the time of triage is essential to develop the most appropriate pain-relieving strategy. The algometric scales that can be used in children in whom self-assessment is not possible are based on the evaluation of behavioral and physiological parameters. However, the actual use of algometric scales in the ED is scarce due to environmental factors, heterogeneity of the scales and lack of training, thus making automated pain assessment desirable. In this study, we propose a camera-based system to provide an objective and contactless pain assessment in children aged less than 3 years, through the automatic detection of behavioral parameters from video recordings. To investigate the feasibility of its usage in the ED environment, we collected video recordings of healthy children aged 3–36 months admitted to the ED with acute pain as the main or accompanying symptom, while pain was measured by a healthcare professional according to the Face, Legs, Activity, Cry, and Consolability (FLACC) pain scale. For the recorded videos, we compared the scores for the items Face (F), Legs (L) and Activity (A) given by the operator with the ones given by our system, analyzing the potentiality and limitations of our approach. By showing that automatic pain assessment in young children in the ED could integrate human evaluation to make it easier and faster, without substituting it, we provide the basis for further research in this field.

Healthcare systems currently store a large amount of clinical data, mostly unstructured textual information, such as electronic health records (EHRs). Manually extracting valuable information from these documents is costly for healthcare professionals. For example, when a patient first arrives at an oncology clinical analysis unit, clinical staff must extract information about the type of neoplasm in order to assign the appropriate clinical specialist. Automating this task is equivalent to text classification in natural language processing (NLP). In this study, we have attempted to extract the neoplasm type by processing Spanish clinical documents. A private corpus of 23, 704 real clinical cases has been processed to extract the three most common types of neoplasms in the Spanish territory: breast, lung and colorectal neoplasms. We have developed methodologies based on state-of-the-art text classification task, strategies based on machine learning and bag-of-words, based on embedding models in a supervised task, and based on bidirectional recurrent neural networks with convolutional layers (C-BiRNN). The results obtained show that the application of NLP methods is extremely helpful in performing the task of neoplasm type extraction. In particular, the 2-BiGRU model with convolutional layer and pre-trained fastText embedding obtained the best performance, with a macro-average, more representative than the micro-average due to the unbalanced data, of 0.981 for precision, 0.984 for recall and 0.982 for F1-score.

A mathematical model of viral infection of a small group of people randomly moving indoors is proposed. The model consists of two different modules. In the first module, based on the standard three-stage cell model, we modified the initial stage of infection. Firstly, the initial degree of immunity is taken into account. Secondly, it is assumed that there is a critical concentration of the pathogen, starting from which there is an active infection of the hosts cells of the human organism. Thirdly, an additional flux of virions from the local atmosphere near the exposed individual is included. The second module presents a model of random movement of people indoors with obstacles. The desired velocity of human consists of a random component and a deterministic velocity that occurs during evacuation from the room. The source of the chaotic desired velocity is modeled by a random Gaussian color process. The actual velocity of individuals is a consequence of the social behavior of the group indoor with obstacles. Physical contacts between individuals and obstacles are described on the basis of effective potential. The numerical study of both the dynamics of the pathogen in an organism in an atmosphere with a random flow of virions and the chaotic movement of individuals indoors is modeled on the basis of a system of stochastic ordinary differential equations.

The COVID-19 pandemic has posed a significant public health challenge on a global scale. It is imperative that we continue to undertake research in order to identify early markers of disease progression, enhance patient care through prompt diagnosis, identification of high-risk patients, early prevention, and efficient allocation of medical resources. In this particular study, we obtained 100 5-min electrocardiograms (ECGs) from 50 COVID-19 volunteers in two different positions, namely upright and supine, who were categorized as either moderately or critically ill. We used classification algorithms to analyze heart rate variability (HRV) metrics derived from the ECGs of the volunteers with the goal of predicting the severity of illness. Our study choose a configuration pro SVC that achieved 76% of accuracy, and 0.84 on F1 Score in predicting the severity of Covid-19 based on HRV metrics.

The effectiveness of confining the population has been observed for centuries. However, this effectiveness has now been demonstrated with data during the COVID-19 pandemic, for which an enormous amount of data and studies are available. In this sense, this paper identifies the determination of the number of people susceptible to contracting the disease, which is present in many epidemic transmission models, as the fundamental variable for understanding the dynamics of infections. We primarily consider the SIRD model, but the data are also contrasted with models and techniques used in pandemic analysis. In addition, the facts and conditions of the COVID-19 pandemic are compared with others that occurred historically.

Inorganic polyphosphates (polyP) consist of linear chains of orthophosphate units linked together by high-energy phosphoanhydride bonds. The family of polyP molecules are evolutionarily old biopolymers and found from bacteria to man. PolyP is exceptional, no other molecule concentrates as much (bio)chemically usable energy as polyP in animals, including humans. Before this discovery, we found that the long-neglected polymer provides orthophosphate units required for bone (hydroxyapatite) synthesis. Hence, polyP is a cornerstone for bone synthesis and repair, especially in higher animals. Besides its importance for regenerative medicine, especially for the reconstitution of osteoarticular impairments/defects, a further imperative property could be attributed the polyP. This polymer is the only extracellular generator of metabolic energy in the form of ATP. While the mitochondria synthesize ATP in large amounts intracellularly, it is polyP, which functions as the storage for extracellular ATP. After enzymatic hydrolysis of polyP by alkaline phosphatase (ALP) the released free energy is partially stored in ADP (formed from AMP), which in the second step is up-phosphorylated to ATP by adenylate kinase (ADK). In turn, the two enzymes ALP and ADK are the biocatalytic proteins that conserve the released free energy and store it in ATP, especially in the extracellular space. In a proof-of-concept, we could demonstrate that polyP is an essential component for human regeneration processes, especially in those regions, which are poorly vascularised, like in bone, cartilage and wounds (including chronic wounds).

EEG is considered an important tool in the diagnostic and treatment process of patients with neurological manifestations of COVID-19, especially with encephalopathy, seizures, and status epilepticus. The present research was aimed at quantitative and visual analysis of the EEG of 85 neuropsychiatric outpatients with COVID-19 history and with psychosomatic complaints at the time of the examination. The control group consisted of 35 healthy subjects. Three types of EEG patterns have been established: polymorphic low-frequency activity; low-frequency polymorphic activity with a predominance of delta, theta rhythms; high frequency EEG with a visible dominant of the beta1 range. The correlation index in the alpha range is stable for the EEG in the control group, where in 90% of the subjects the correlation coefficients in the alpha range were more than 0.6. On the contrary, patients have a polymorphic picture, stable indicators with a coefficient of more than 0.6 for all the studied connections, both between the hemispheres and within the hemispheres were registered only in 25% of the subjects. Analysis of the coherence coefficients in patients, on the contrary, shows a higher stability of interhemispheric connections and various options for reducing connections within the hemispheres, which often have a “mirror character”.