Local Translation Research Articles

Investigating two-wheelers risk factors for severe crashes using an interpretable machine learning approach and SHAP analysis

The use of two-wheelers (TWs) has gained popularity as an alternative to personal vehicles due to their flexibility, fuel economy, ease of parking, and size, especially in congested cities. However, TWs are considered vulnerable road users due to their higher riding risk compared to other modes. This study proposes a novel framework to extract latent and dependent heterogeneous risk factors that affect the crash severity of TWs. By combining eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanation (SHAP) analysis, this study investigates the factors affecting TW crash severity, providing both local and global interpretability. The XGBoost method is employed to model crash severity, while SHAP analysis facilitates the derivation of explanations from the model, enhancing our understanding of the contributing factors. The French crash dataset for TWs between 2014 and 2017 is utilized for this analysis. The findings highlight that the department of the crash, road category, urbanization level, TW category, and age of the user significantly influence TW crash severity. Furthermore, severe injuries are more likely to occur in TW crashes associated with rural areas, older riders, riders not wearing helmets, run-off-road crashes, and crossing roads. The insights derived from this study can be leveraged to develop targeted interventions that address the identified risk factors and promote the safety of TW riders. By focusing on these key factors, policymakers and stakeholders can implement effective measures to reduce the severity of TW crashes and enhance the overall safety of TW users.

Accurate, interpretable predictions of materials properties within transformer language models

Property prediction accuracy has long been a key parameter of machine learning in materials informatics. Accordingly, advanced models showing state-of-the-art performance turn into highly parameterized black boxes missing interpretability. Here, we present an elegant way to make their reasoning transparent. Human-readable text-based descriptions automatically generated within a suite of open-source tools are proposed as materials representation. Transformer language models pretrained on 2 million peer-reviewed articles take as input well-known terms such as chemical composition, crystal symmetry, and site geometry. Our approach outperforms crystal graph networks by classifying four out of five analyzed properties if one considers all available reference data. Moreover, fine-tuned text-based models show high accuracy in the ultra-small data limit. Explanations of their internal machinery are produced using local interpretability techniques and are faithful and consistent with domain expert rationales. This language-centric framework makes accurate property predictions accessible to people without artificial-intelligence expertise.

In situ sequencing of localized translation.

In situ sequencing of localized translation.

Understanding Arteriosclerotic Heart Disease Patients Using Electronic Health Records: A Machine Learning and Shapley Additive exPlanations Approach.

The number of deaths from cardiovascular disease is projected to reach 23.3 million by 2030. As a contribution to preventing this phenomenon, this paper proposed a machine learning (ML) model to predict patients with arteriosclerotic heart disease (AHD). We also interpreted the prediction model results based on the ML approach and deployed modelagnostic ML methods to identify informative features and their interpretations. We used a hematology Electronic Health Record (EHR) with information on erythrocytes, hematocrit, hemoglobin, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, leukocytes, thrombocytes, age, and sex. To detect and predict AHD, we explored random forest (RF), XGBoost, and AdaBoost models. We examined the prediction model results based on the confusion matrix and accuracy measures. We used the Shapley Additive exPlanations (SHAP) framework to interpret the ML model and quantify the contribution of features to predictions. Our study included data from 6,837 patients, with 4,702 records from patients diagnosed with AHD and 2,135 records from patients without an AHD diagnosis. AdaBoost outperformed RF and XGBoost, achieving an accuracy of 0.78, precision of 0.82, F1-score of 0.85, and recall of 0.88. According to the SHAP summary bar plot method, hemoglobin was the most important attribute for detecting and predicting AHD patients. The SHAP local interpretability bar plot revealed that hemoglobin and mean corpuscular hemoglobin concentration had positive impacts on AHD prediction based on a single observation. ML models based on real clinical data can be used to predict AHD.

Specific-Input LIME Explanations for Tabular Data Based on Deep Learning Models

Deep learning researchers believe that as deep learning models evolve, they can perform well on many tasks. However, the complex parameters of deep learning models make it difficult for users to understand how deep learning models make predictions. In this paper, we propose the specific-input local interpretable model-agnostic explanations (LIME) model, a novel interpretable artificial intelligence (XAI) method that interprets deep learning models of tabular data. The specific-input process uses feature importance and partial dependency plots (PDPs) to select the “what” and “how”. In our experiments, we first obtain a basic interpretation of the data by simulating user behaviour. Second, we use our approach to understand “which” features deep learning models focus on and how these features affect the model’s predictions. From the experimental results, we find that this approach improves the stability of LIME interpretations, compensates for the problem of LIME only focusing on local interpretations, and achieves a balance between global and local interpretations.

PTBP1 regulates injury responses and sensory pathways in adult peripheral neurons.

Polypyrimidine tract binding protein 1 (PTBP1) is thought to be expressed only at embryonic stages in central neurons. Its down-regulation triggers neuronal differentiation in precursor and non-neuronal cells, an approach recently tested for generation of neurons de novo for amelioration of neurodegenerative disorders. Moreover, PTBP1 is replaced by its paralog PTBP2 in mature central neurons. Unexpectedly, we found that both proteins are coexpressed in adult sensory and motor neurons, with PTBP2 restricted mainly to the nucleus, while PTBP1 also shows axonal localization. Levels of axonal PTBP1 increased markedly after peripheral nerve injury, and it associates in axons with mRNAs involved in injury responses and nerve regeneration, including importin β1 (KPNB1) and RHOA. Perturbation of PTBP1 affects local translation in axons, nociceptor neuron regeneration and both thermal and mechanical sensation. Thus, PTBP1 has functional roles in adult axons. Hence, caution is required before considering targeting of PTBP1 for therapeutic purposes.

Interpreting the decisions of CNNs via influence functions.

An understanding of deep neural network decisions is based on the interpretability of model, which provides explanations that are understandable to human beings and helps avoid biases in model predictions. This study investigates and interprets the model output based on images from the training dataset, i.e., to debug the results of a network model in relation to the training dataset. Our objective was to understand the behavior (specifically, class prediction) of deep learning models through the analysis of perturbations of the loss functions. We calculated influence scores for the VGG16 network at different hidden layers across three types of disturbances in the original images of the ImageNet dataset: texture, style, and background elimination. The global and layer-wise influence scores allowed the identification of the most influential training images for the given testing set. We illustrated our findings using influence scores by highlighting the types of disturbances that bias predictions of the network. According to our results, layer-wise influence analysis pairs well with local interpretability methods such as Shapley values to demonstrate significant differences between disturbed image subgroups. Particularly in an image classification task, our layer-wise interpretability approach plays a pivotal role to identify the classification bias in pre-trained convolutional neural networks, thus, providing useful insights to retrain specific hidden layers.

Cytosolic and mitochondrial translation elongation are coordinated through the molecular chaperone TRAP1 for the synthesis and import of mitochondrial proteins.

A complex interplay between mRNA translation and cellular respiration has been recently unveiled, but its regulation in humans is poorly characterized in either health or disease. Cancer cells radically reshape both biosynthetic and bioenergetic pathways to sustain their aberrant growth rates. In this regard, we have shown that the molecular chaperone TRAP1 not only regulates the activity of respiratory complexes, behaving alternatively as an oncogene or a tumor suppressor, but also plays a concomitant moonlighting function in mRNA translation regulation. Herein, we identify the molecular mechanisms involved, showing that TRAP1 (1) binds both mitochondrial and cytosolic ribosomes, as well as translation elongation factors; (2) slows down translation elongation rate; and (3) favors localized translation in the proximity of mitochondria. We also provide evidence that TRAP1 is coexpressed in human tissues with the mitochondrial translational machinery, which is responsible for the synthesis of respiratory complex proteins. Altogether, our results show an unprecedented level of complexity in the regulation of cancer cell metabolism, strongly suggesting the existence of a tight feedback loop between protein synthesis and energy metabolism, based on the demonstration that a single molecular chaperone plays a role in both mitochondrial and cytosolic translation, as well as in mitochondrial respiration.

Benchmarking EEG-based Cross-dataset Driver Drowsiness Recognition with Deep Transfer Learning.

It usually takes a long time to collect data for calibration when using electroencephalography (EEG) for driver drowsiness monitoring. Cross-dataset recognition is desirable since it can significantly save the calibration time when an existing dataset is used. However, the recognition accuracy is affected by the distribution drift problem caused by different experimental environments when building different datasets. In order to solve the problem, we propose a deep transfer learning model named Entropy-Driven Joint Adaptation Network (EDJAN), which can learn useful information from source and target domains simultaneously. An entropy-driven loss function is used to promote clustering of target-domain representations and an individual-level domain adaptation technique is proposed to alleviate the distribution discrepancy problem of test subjects. We use two public driving datasets SEEG-VIG and SADT to test the model on the cross-dataset setting. The proposed model achieved an accuracy of 83.3% when SADT is used as source domain and SEED-VIG is used as target domain and 76.7% accuracy on the reverse setting, which is higher than the other SOTA methods. The results are further analyzed with both global and local interpretation methods. Our work illuminates a promising direction of using EEG for calibration-free driver drowsiness recognition.

Garden-path sentences and the diversity of their (mis)representations.

Previous studies have reliably shown that the initial misanalysis of garden-path sentences lingers even after the whole sentence is processed. However, other aspects of the resulting representation of these sentences are far from being clear. Some authors argue that comprehenders form a full analysis of the sentence which is faithful to the input and that the fact that the misanalysis lingers is due to an inhibition failure. Recently, it has been shown that comprehenders might not manage to create a coherent representation at all, at least in the case of more demanding garden-path structures. The aim of the current paper is to examine resulting representations of garden-path sentences in more detail. To do this, four self-paced reading experiments in Czech were conducted, which differed in the presentation mode (word-by-word and sentence-at-once) and comprehension question format (yes-no questions and open-ended questions). The experiments replicated effects typical for the lingering initial misanalysis, but provided mixed evidence for other aspects of resulting representations. In most cases, participants managed to build a coherent representation that was faithful to the input. However, both the quantitative and qualitative analysis of the results showed that comprehenders sometimes maintained multiple local interpretations at once or even failed to build a coherent representation of a garden-path sentence. Thus, we argue that resulting representations of garden-path sentences are in fact not uniform, but rather diverse, and they vary both in their faithfulness to the presented input and in their internal coherence.

Predicting Depression during the COVID-19 Pandemic Using Interpretable TabNet: A Case Study in South Korea

COVID-19 has further aggravated problems by compelling people to stay indoors and limit social interactions, leading to a worsening of the depression situation. This study aimed to construct a TabNet model combined with SHapley Additive exPlanations (SHAP) to predict depression in South Korean society during the COVID-19 pandemic. We used a tabular dataset extracted from the Seoul Welfare Survey with a total of 3027 samples. The TabNet model was trained on this dataset, and its performance was compared to that of several other machine learning models, including Random Forest, eXtreme Gradient Boosting, Light Gradient Boosting, and CatBoost. According to the results, the TabNet model achieved an Area under the receiver operating characteristic curve value (AUC) of 0.9957 on the training set and an AUC of 0.9937 on the test set. Additionally, the study investigated the TabNet model’s local interpretability using SHapley Additive exPlanations (SHAP) to provide post hoc global and local explanations for the proposed model. By combining the TabNet model with SHAP, our proposed model might offer a valuable tool for professionals in social fields, and psychologists without expert knowledge in the field of data analysis can easily comprehend the decision-making process of this AI model.

Creative Kampong and the Promotion of Sustainable Development Knowledge in Urban Grassroots: The Case of Surakarta

As the global trend is going into circular economy paradigm. The trend encourages business entities to focus on the production of durable, efficient, and reusable commodities to promote sustainable development. On the demand side, the market also gradually adjusts its preferences to commodities with sustainability values. With limited scale of capital, networks, and technology, Small and Medium Enterprises (SMEs) are usually lagged behind in catching new knowledge, especially those exchanged internationally. Sustainability, which is as an important knowledge under the circular economy trend, has caused serious problem for SMEs because those unable to catch the knowledge will be incompetent to serve the new market preferences and vulnerable to survive. Urban planning practices introduce the business district as one of spatial concepts to mediate the problem. It encourages enterprises with business linkages to locate next to each other and share activities and facilities. In Indonesia, the creative kampong is a local translation of business district concept, which emerges as a national hype. This research aims at studying the empirical case of the creative kampong and its knowledge transfer effectiveness for SMEs. By using the logic of Actor Network Theory (ANT) and focusing on the case study of four creative kampongs in Surakarta, we unravel the level of understanding of SMEs towards sustainability. Our data are based on stakeholder interviews and documentary reviews. The findings reveal that the majority of SMEs were understand of sustainability, and the creative kampong provided a significant contribution to ease the transfer knowledge process.

Local coordination of mRNA storage and degradation near mitochondria modulates C. elegans ageing

Mitochondria are central regulators of healthspan and lifespan, yet the intricate choreography of multiple, tightly controlled steps regulating mitochondrial biogenesis remains poorly understood. Here, we uncover a pivotal role for specific elements of the 5′‐3′ mRNA degradation pathway in the regulation of mitochondrial abundance and function. We find that the mRNA degradation and the poly‐A tail deadenylase CCR4‐NOT complexes form distinct foci in somatic Caenorhabditis elegans cells that physically and functionally associate with mitochondria. Components of these two multi‐subunit complexes bind transcripts of nuclear‐encoded mitochondria‐targeted proteins to regulate mitochondrial biogenesis during ageing in an opposite manner. In addition, we show that balanced degradation and storage of mitochondria‐targeted protein mRNAs are critical for mitochondrial homeostasis, stress resistance and longevity. Our findings reveal a multifaceted role of mRNA metabolism in mitochondrial biogenesis and show that fine‐tuning of mRNA turnover and local translation control mitochondrial abundance and promote longevity in response to stress and during ageing.

Endogenous epitope tagging of eEF1A2 in mice reveals early embryonic expression of eEF1A2 and subcellular compartmentalisation of neuronal eEF1A1 and eEF1A2

All vertebrate species express two independently-encoded forms of translation elongation factor eEF1A. In humans and mice eEF1A1 and eEF1A2 are 92 % identical at the amino acid level, but the well conserved developmental switch between the two variants in specific tissues suggests the existence of important functional differences. Heterozygous mutations in eEF1A2 result in neurodevelopmental disorders in humans; the mechanism of pathogenicity is unclear, but one hypothesis is that there is a dominant negative effect on eEF1A1 during development. The high degree of similarity between the eEF1A proteins has complicated expression analysis in the past; here we describe a gene edited mouse line in which we have introduced a V5 tag in the gene encoding eEF1A2. Expression analysis using anti-V5 and anti-eEF1A1 antibodies demonstrates that, in contrast to the prevailing view that eEF1A2 is only expressed postnatally, it is expressed from as early as E11.5 in the developing neural tube. Two colour immunofluorescence also reveals coordinated switching between eEF1A1 and eEF1A2 in different regions of postnatal brain. Completely reciprocal expression of the two variants is seen in post-weaning mouse brain with eEF1A1 expressed in oligodendrocytes and astrocytes and eEF1A2 in neuronal soma. Although eEF1A1 is absent from neuronal cell bodies after development, it is widely expressed in axons. This expression does not appear to coincide with myelin sheaths originating from oligodendrocytes but rather results from localised translation within the axon, suggesting that both variants are transcribed in neurons but show completely distinct subcellular localisation at the protein level. These findings will form an underlying framework for understanding how missense mutations in eEF1A2 result in neurodevelopmental disorders.

DT-CTNet: A clinically interpretable diagnosis model for fetal distress

Clinically, Fetal Heart Rate (FHR)-based intelligent cardiotocography classification to diagnose fetal well-being is of utmost importance for obstetricians and gynecologists. However, the current Fetal Distress Diagnosis (FDD) algorithms based on Artificial Intelligence (AI) methods are designed as black box models that usually result in poor performance on clinical interpretation and lack explicit diagnostic evidence. This paper focuses on two aspects: enhancing the accuracy of FD and providing explanations for the model’s decisions. Specifically, we propose a novel clinically interpretable dual-stream AI architecture for FDD. The two streams embed multi-feature category representations and the original FHR into the Digital Twin model (DT) and the deep convolutional neural Networks-based Case Tracking model (CTNet), respectively, thus marking as DT-CTNet. And the former makes the input-outcome dependency transparent, while the latter provides the samples that are the most similar to the object being diagnosed. By this design, not only is high-precision FDD achievable, but a transparent explanation with global, local and instantiation interpretation can be obtained. To the best of our knowledge, this work is the first to provide an interpretable diagnosis of fetal distress. The proposed method is comprehensively evaluated from five aspects: classification model accuracy, explanation of the classification model decision, consistency of the evidence explanation, pathological difference in visual analysis, and a comparison with state-of-the-art methods. Experimental results, which are performed on a challenging dataset, show that the proposed fetal distress diagnosis method outperforms the state-of-the-art methods and has a high accuracy (0.963).

Effect of Cilostazol on Delayed Cerebral Infarction in Aneurysmal Subarachnoid Hemorrhage Using Explainable Predictive Modeling.

The studies interpreting DCI, a complication of SAH, and identifying correlations are very limited. This study aimed to investigate the effect of cilostazol on ACV and DCI after coil embolization for ruptured aneurysms (n = 432). A multivariate analysis was performed and explainable artificial intelligence approaches were used to analyze the contribution of cilostazol as a risk factor on the development of ACV and DCI with respect to global and local interpretation. The cilonimo group was significantly lower than the nimo group in ACV (13.5% vs. 29.3; p = 0.003) and DCI (7.9% vs. 20.7%; p = 0.006), respectively. In a multivariate logistic regression, the odds ratio for DCI for the cilonimo group, female sex, and aneurysm size was 0.556 (95% confidence interval (CI), 0.351-0.879; p = 0.012), 3.713 (95% CI, 1.683-8.191; p = 0.001), and 1.106 (95% CI, 1.008-1.214; p = 0.034). The risk of a DCI occurrence was significantly increased with an aneurysm size greater than 10 mm (max 80%). The mean AUC of the XGBoost and logistic regression models was 0.94 ± 0.03 and 0.95 ± 0.04, respectively. Cilostazol treatment combined with nimodipine could decrease the prevalence of ACV (13.5%) and DCI (7.9%) in patients with aSAH.

Eksistensi, Regulasi, dan Fungsi Meunasah Sebagai Lembaga Pendidikan Islam di Aceh

This article aims to analyze the existence, regulation, and function of the meunasah as a center for religious and cultural activities in society, including Islamic religious education institutions. This study uses a qualitative approach. Data were collected from literature related to research objects and field data through observation, interviews and documentation. The research findings show that the existence of meunasah is not only a cultural heritage which is the center of community activities, but also an institution that has a clear legal standing. The clarity of the legal status of the meunasah as a center for various community activities departs from the special status of Aceh as an autonomous region. The existence of meunasah is a local translation for educational activities in the Muslim world community, especially from the Middle East, namely madrasas. The word meunasah is also an adaptation of the tongue of the Acehnese people from the word madrasa. By regulation, various qanuns have been stipulated regarding the existence of the meunasah. Among the important functions of the meunasah is the implementation of Islamic religious education. The existence of religious education institutions in meunasah in agrarian societies is part of the hierarchy of the traditional education system. Meanwhile in urban areas, the existence of meunasah as educational institutions is a center of recitation such as center for religious studies and center for the study of the Qur’an.

Spatially resolved single-cell translatomics at molecular resolution.

The precise control of messenger RNA (mRNA) translation is a crucial step in posttranscriptional gene regulation of cellular physiology. However, it remains a challenge to systematically study mRNA translation at the transcriptomic scale with spatial and single-cell resolution. Here, we report the development of ribosome-bound mRNA mapping (RIBOmap), a highly multiplexed three-dimensional in situ profiling method to detect cellular translatome. RIBOmap profiling of 981 genes in HeLa cells revealed cell cycle-dependent translational control and colocalized translation of functional gene modules. We mapped 5413 genes in mouse brain tissues, yielding spatially resolved single-cell translatomic profiles for 119,173 cells and revealing cell type-specific and brain region-specific translational regulation, including translation remodeling during oligodendrocyte maturation. Our method detected widespread patterns of localized translation in neuronal and glial cells in intact brain tissue networks.

Geochemical characteristics and oil–oil correlation of the upper Cretaceous oils from the Iranian part of the Persian Gulf Basin

The Cenomanian Sarvak oil reservoirs are distributed over large areas of the Persian Gulf basin. The purpose of this study is analyzing the geochemical characteristics of the Sarvak oil reservoirs and their inter-relationships in the Persian Gulf, classification of the Sarvak oil samples and investigation of the possible causes of the genetic difference in oil families. In the previous studies, limited samples of Sarvak oil reservoir in scarce oilfields were studied individually and local interpretations are made accordingly. The current study employs a more complete set of geochemical from the Iranian part of Persian Gulf and regional interpretations are drawn. To achieve this goal, the geochemical data of 41 oil samples from 10 oilfields were collected and assessed based on gas chromatography (GC), gas chromatography–mass spectrometry (GC–MS), and stable carbon isotope analysis. It was demonstrated with the evaluations that the oils accumulated in the Upper Cretaceous Sarvak reservoirs in the Persian Gulf basin originate from different source rocks. The oil samples are genetically classified into four oil families based on the similarities and differences of parameters related to the depositional environment-dependent parameters using hierarchical cluster analysis (HCA), star diagram, and stable carbon isotope diagram. The source rocks were mainly deposited in anoxic marine carbonate environments. The thermal maturity of the oils was evaluated using steranes and trisnorhopanes. Oil families 2 and 4 (located at the center of the Persian Gulf) have the highest thermal maturity compared to the other samples; in contrast, oil family 1 (located at the west of the Persian Gulf) has the lowest thermal maturity. The possible source rocks of oil family 1 and family 3 (located at eastern Persian Gulf) using C28/C29 steranes are the upper Cretaceous successions; in contrast, the possible source rocks of oil families 2 and 4 are the upper Triassic and Lower Jurassic formations. Results of the study show a high similarity between the oils of families 2 and 4, which are located at the center of the Persian Gulf; these results also recognize the significant difference between these two oil families and oil families 1 and 3. The significant issue raised in this study is to find out the reason for this difference. The structural analysis of the central Persian Gulf shows that huge vertical faults created by salt domes intrusion provided the migration pathways for trapping oil in Sarvak reservoirs. The oils of the Surmeh, Fahliyan, and Dariyan reservoirs have migrated upwards through these faults to be trapped in the Sarvak Formation across the central Persian Gulf.

3'UTR of mRNA Encoding CPEB Protein Orb2 Plays an Essential Role in Intracellular Transport in Neurons.

Intracellular trafficking plays a critical role in the functioning of highly polarized cells, such as neurons. Transport of mRNAs, proteins, and other molecules to synaptic terminals maintains contact between neurons and ensures the transmission of nerve impulses. Cytoplasmic polyadenylation element binding (CPEB) proteins play an essential role in long-term memory (LTM) formation by regulating local translation in synapses. Here, we show that the 3'UTR of the Drosophila CPEB gene orb2 is required for targeting the orb2 mRNA and protein to synapses and that this localization is important for LTM formation. When the orb2 3'UTR is deleted, the orb2 mRNAs and proteins fail to localize in synaptic fractions, and pronounced LTM deficits arise. We found that the phenotypic effects of the orb2 3'UTR deletion were rescued by introducing the 3'UTR from the orb, another Drosophila CPEB gene. In contrast, the phenotypic effects of the 3'UTR deletion were not rescued by the 3'UTR from one of the Drosophila α-tubulin genes. Our results show that the orb2 mRNAs must be targeted to the correct locations in neurons and that proper targeting depends upon sequences in the 3'UTR.