Exploring dimension-reduced embeddings with Sleepwalk.
Ontology highlight
ABSTRACT:
SUBMITTER: Ovchinnikova S
PROVIDER: S-EPMC7263188 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:BackgroundWord embeddings are dense numeric vectors used to represent language in neural networks. Until recently, there had been no publicly released embeddings trained on clinical data. Our work is the first to study the privacy implications of releasing these models.ObjectiveThis paper aims to demonstrate that traditional word embeddings created on clinical corpora that have been deidentified by removing personal health information (PHI) can nonetheless be exploited to reveal sensitive patient information.MethodsWe used embeddings created from 400,000 doctor-written consultation notes and experimented with 3 common word embedding methods to explore the privacy-preserving properties of each.ResultsWe found that if publicly released embeddings are trained from a corpus anonymized by PHI removal, it is possible to reconstruct up to 68.5% (n=411/600) of the full names that remain in the deidentified corpus and associated sensitive information to specific patients in the corpus from which the embeddings were created. We also found that the distance between the word vector representation of a patient's name and a diagnostic billing code is informative and differs significantly from the distance between the name and a code not billed for that patient.ConclusionsSpecial care must be taken when sharing word embeddings created from clinical texts, as current approaches may compromise patient privacy. If PHI removal is used for anonymization before traditional word embeddings are trained, it is possible to attribute sensitive information to patients who have not been fully deidentified by the (necessarily imperfect) removal algorithms. A promising alternative (ie, anonymization by PHI replacement) may avoid these flaws. Our results are timely and critical, as an increasing number of researchers are pushing for publicly available health data.
| S-EPMC7391163 | biostudies-literature
Project description:When addressing spatial biological questions using mathematical models, symmetries within the system are often exploited to simplify the problem by reducing its physical dimension. In a reduced-dimension model molecular movement is restricted to the reduced dimension, changing the nature of molecular movement. This change in molecular movement can lead to quantitatively and even qualitatively different results in the full and reduced systems. Within this manuscript we discuss the condition under which restricted molecular movement in reduced-dimension models accurately approximates molecular movement in the full system. For those systems which do not satisfy the condition, we present a general method for approximating unrestricted molecular movement in reduced-dimension models. We will derive a mathematically robust, finite difference method for solving the 2D diffusion equation within a 1D reduced-dimension model. The methods described here can be used to improve the accuracy of many reduced-dimension models while retaining benefits of system simplification.
| S-EPMC8184792 | biostudies-literature
Project description:Inductive proximity sensors (IPSs) present a unique no-contact advantage. They are widely preferred for displacement measurement in various industrial fields (e.g., aviation and aerospace), and they are improved continuously. When the inductance and resistance components of the IPS sensing core are separated, the influence of temperature drift on measurement can be eliminated. The complexity of online computation of component separation can be reduced using a two-dimensional look-up table method. However, this method exhibits disadvantages, such as large capacity of the look-up table, dependency on precision measurement of sensing core parameter, and nonlinear distribution of measurement resolution. This study aims to overcome these disadvantages by examining the nonlinear relationship between the response of the sensing core and the ambient temperature, and proposes a dimension-reduced measurement method. The proposed method extracts the characteristics of the response curves at two temperatures and calculates the characteristics of the response curves at any temperature using a linear approximation. The look-up table capacity is less than 0.37% of the two-dimensional look-up table capacity (condensed) under the same condition; dimension reduction enables the construction of a complete look-up table directly by calibration procedures and avoids precise measurement on sensing core parameters; the calibration procedures establish uniform mapping of the distribution of measurement resolution. The experiment shows that, when the measurement ranges are 0-6, 0-5, and 0-4 mm, the maximum measurement errors are 0.140, 0.065, and 0.040 mm, respectively, under temperature ranging from 20 ∘ C to 110 ∘ C. This study extends the measurement range from 0-5 mm to 0-7 mm and improves the measurement accuracy over 0.1 mm (50% at 5 mm) compared with the two-dimensional look-up table method. Therefore, the proposed method not only inherits the advantages of the original method but also achieves the above-mentioned expected capacity improvements effectively.
| S-EPMC5539794 | biostudies-other
Project description:Background: Neuronal activity can be modeled as a nonlinear dynamical system to yield novel measures of neuronal state and dysfunction. We hypothesized that electrical activity during neuronal differentiation would be marked by a reduction in dynamical complexity in autism spectrum disorder (ASD). Methods: Electrical activity of induced pluripotent stem cell (iPSC)-derived neurons from ASD patients and controls were recorded using a multielectrode array (MEA). Minimum embedding dimension (MED) analysis was performed to characterize the complexity present in the electrical recordings of the neuronal cultures during differentiation. Gene expression changes with respect to the MED were analyzed, and gene ontology analysis and the BrainSpan Atlas were used to identify gene enrichments in specific biological pathways, brain regions, and developmental stages. Results: We found that the MED showed a statistically significant deficit in the ASD samples, suggesting that, compared to controls, network complexity is reduced in ASD during neuronal differentiation. MED was correlated to clinical endpoints such as nonverbal intelligence and was associated with 53 differentially expressed genes. These differentially expressed genes were overrepresented in ASD gene lists and pathways related to neurodevelopment, cell morphology, and cell migration. Spatiotemporal analysis also showed a prenatal temporal enrichment in cortical and deep brain structures. Conclusion: Reduced network complexity in ASD was observed by MED analysis during neuronal differentiation, indicating deficits in integrated and synchronized firing. The results present nonlinear dynamical analysis of iPSC-derived neuronal electrical recordings as a new paradigm for the investigation of aberrant brain activity in neurodevelopmental disorders.
2020-05-18 | GSE125020 | GEO
Project description:Objectives. To evaluate the impact of adding a respiratory dimension (a bolt-on dimension) to the EQ-5D-5L health state valuations. Methods. Based on extensive regression and principal component analyses, 2 respiratory bolt-on candidates were formulated: R1, limitations in physical activities due to shortness of breath, and R2, breathing problems. Valuation interviews for the selected bolt-ons were performed with a representative sample from the Dutch general public using the standardized interview protocol and software of the EuroQol group. Hybrid models based on the combined time-tradeoff (TTO) and discrete choice experiment (DCE) data were estimated to assess whether the 5 levels of the respiratory bolt-on led to significant changes in utility values. Results. For each bolt-on candidate, slightly more than 200 valuation interviews were conducted. Mean TTO values and DCE choice probabilities for health states with a level 4 or 5 for the respiratory dimension were significantly lower compared with the same health states in the Dutch EQ-5D-5L valuation study without the respiratory dimension. Results of hybrid models showed that for the bolt-on "limitations in physical activities," the utility decrements were significant for level 3 (-0.055), level 4 (-0.087), and level 5 (-0.135). For "breathing problems," the utility decrements for the same levels were greater (-0.086, -0.219, and -0.327, respectively). Conclusions. The addition of each of the 2 respiratory bolt-ons to the EQ-5D-5L had a significant effect on the valuation of health states with severe levels for the bolt-on. The bolt-on dimension "breathing problems" showed the greatest utility decrements and therefore seems the most appropriate respiratory bolt-on dimension.
| S-EPMC6613181 | biostudies-literature
Project description:Antibiotic use in livestock production is one of the drivers of antibiotic resistance and a shift towards better and reduced antibiotic usage is urgently required. In Vietnam, where there are frequent reports of the misuse and overuse of antibiotics, little attention has been paid to farmers who have successfully changed their practices. This qualitative study aims to understand the transition process of Vietnamese chicken farmers toward reduced antibiotic usage. We conducted semi-structured interviews with 18 chicken farmers, 13 drug sellers, and 5 traders using participatory tools and a socio-anthropological approach. We explored the farmers' histories, current and past antibiotic usage, methods used to reduce antibiotic use, and motivations and barriers to changing practices. Through the thematic analysis of the farmers' transcripts, we identified technical, economic, and social factors that influence change. Out of eighteen farmers, we identified ten farmers who had already reduced antibiotic usage. The main motivations included producing quality chickens (tasty and safe) while reducing farm expenditures. Barriers were related to poor biosecurity in the area, market failures, and the farmers' lack of knowledge. Innovation led to overcome these obstacles included the local development of handmade probiotics and the organization of farmer cooperatives to overcome economic difficulties and guarantee product outlets. Knowledge was increased by workshops organized at the communal level and the influence of competent veterinarians in the area. We showed that the transition process was influenced by several components of the system rather than by any individual alone. Our study demonstrated that local initiatives to reduce antibiotic use in Vietnamese chicken production do exist. As changes depend on the system in which stakeholders are embedded, systemic lock-ins must be removed to allow practices to change. The promotion of locally-developed solutions should be further encouraged.
| S-EPMC10490891 | biostudies-literature
Project description:The brain predicts the sensory consequences of our movements and uses these predictions to attenuate the perception of self-generated sensations. Accordingly, self-generated touch feels weaker than an externally generated touch of identical intensity. In schizophrenia, this somatosensory attenuation is substantially reduced, suggesting that patients with positive symptoms fail to accurately predict and process self-generated touch. If an impaired prediction underlies the positive symptoms of schizophrenia, then a similar impairment should exist in healthy nonclinical individuals with high positive schizotypal traits. One hundred healthy participants (53 female), assessed for schizotypal traits, underwent a well-established psychophysics force discrimination task to quantify how they perceived self-generated and externally generated touch. The perceived intensity of tactile stimuli delivered to their left index finger (magnitude) and the ability to discriminate the stimuli (precision) was measured. We observed that higher positive schizotypal traits were associated with reduced somatosensory attenuation and poorer somatosensory precision of self-generated touch, both when treating schizotypy as a continuous or categorical variable. These effects were specific to positive schizotypy and were not observed for the negative or disorganized dimensions of schizotypy. The results suggest that positive schizotypal traits are associated with a reduced ability to predict and process self-generated touch. Given that the positive dimension of schizotypy represents the analogue of positive psychotic symptoms of schizophrenia, deficits in processing self-generated tactile information could indicate increased liability to schizophrenia.
| S-EPMC9261081 | biostudies-literature
Project description:The study of non-human animals' communication systems generally relies on the transcription of vocal sequences using a finite set of discrete units. This set is referred to as a vocal repertoire, which is specific to a species or a sub-group of a species. When conducted by human experts, the formal description of vocal repertoires can be laborious and/or biased. This motivates computerised assistance for this procedure, for which machine learning algorithms represent a good opportunity. Unsupervised clustering algorithms are suited for grouping close points together, provided a relevant representation. This paper therefore studies a new method for encoding vocalisations, allowing for automatic clustering to alleviate vocal repertoire characterisation. Borrowing from deep representation learning, we use a convolutional auto-encoder network to learn an abstract representation of vocalisations. We report on the quality of the learnt representation, as well as of state of the art methods, by quantifying their agreement with expert labelled vocalisation types from 8 datasets of other studies across 6 species (birds and marine mammals). With this benchmark, we demonstrate that using auto-encoders improves the relevance of vocalisation representation which serves repertoire characterisation using a very limited number of settings. We also publish a Python package for the bioacoustic community to train their own vocalisation auto-encoders or use a pretrained encoder to browse vocal repertoires and ease unit wise annotation.
| S-EPMC10332598 | biostudies-literature
Project description:MotivationMachine-learning models trained on protein sequences and their measured functions can infer biological properties of unseen sequences without requiring an understanding of the underlying physical or biological mechanisms. Such models enable the prediction and discovery of sequences with optimal properties. Machine-learning models generally require that their inputs be vectors, and the conversion from a protein sequence to a vector representation affects the model's ability to learn. We propose to learn embedded representations of protein sequences that take advantage of the vast quantity of unmeasured protein sequence data available. These embeddings are low-dimensional and can greatly simplify downstream modeling.ResultsThe predictive power of Gaussian process models trained using embeddings is comparable to those trained on existing representations, which suggests that embeddings enable accurate predictions despite having orders of magnitude fewer dimensions. Moreover, embeddings are simpler to obtain because they do not require alignments, structural data, or selection of informative amino-acid properties. Visualizing the embedding vectors shows meaningful relationships between the embedded proteins are captured.Availability and implementationThe embedding vectors and code to reproduce the results are available at https://github.com/fhalab/embeddings_reproduction/.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC6061698 | biostudies-literature
Project description:MotivationDespite recent advancements in sequencing technologies and assembly methods, obtaining high-quality microbial genomes from metagenomic samples is still not a trivial task. Current metagenomic binners do not take full advantage of assembly graphs and are not optimized for long-read assemblies. Deep graph learning algorithms have been proposed in other fields to deal with complex graph data structures. The graph structure generated during the assembly process could be integrated with contig features to obtain better bins with deep learning.ResultsWe propose GraphMB, which uses graph neural networks to incorporate the assembly graph into the binning process. We test GraphMB on long-read datasets of different complexities, and compare the performance with other binners in terms of the number of High Quality (HQ) genome bins obtained. With our approach, we were able to obtain unique bins on all real datasets, and obtain more bins on most datasets. In particular, we obtained on average 17.5% more HQ bins when compared with state-of-the-art binners and 13.7% when aggregating the results of our binner with the others. These results indicate that a deep learning model can integrate contig-specific and graph-structure information to improve metagenomic binning.Availability and implementationGraphMB is available from https://github.com/MicrobialDarkMatter/GraphMB.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC9525014 | biostudies-literature