Quantification Of Similarity Research Articles

Extended many-item similarity indices for sets of nucleotide and protein sequences

Quantification of similarities between protein sequences or DNA/RNA strands is a (sub-)task that is ubiquitously present in bioinformatics workflows, and is usually accomplished by pairwise comparisons of sequences, utilizing simple (e.g. percent identity) or more intricate concepts (e.g. substitution scoring matrices). Complex tasks (such as clustering) rely on a large number of pairwise comparisons under the hood, instead of a direct quantification of set similarities. Based on our recently introduced framework that enables multiple comparisons of binary molecular fingerprints (i.e., direct calculation of the similarity of fingerprint sets), here we introduce novel symmetric similarity indices for analogous calculations on sets of character sequences with more than two (t) possible items (e.g. DNA/RNA sequences with t = 4, or protein sequences with t = 20). The features of these new indices are studied in detail with analysis of variance (ANOVA), and demonstrated with three case studies of protein/DNA sequences with varying degrees of similarity (or evolutionary proximity). The Python code for the extended many-item similarity indices is publicly available at: https://github.com/ramirandaq/tn_Comparisons.

SLID: Exploiting Spatial Locality in Input Data as a Computational Reuse Method for Efficient CNN

Convolutional Neural Networks (CNNs) revolutionized computer vision and reached the state-of-the-art performance for image processing, object recognition, and video classification. Even though CNN inference is notoriously compute-intensive, as convolutions account for >90% of the total operation tasks, the ability to tradeoff between accuracy, performance, power, and latency to meet target application makes it an open research topic. This paper proposes the Spatial Locality Input Data (SLID) method for computational reuse during the inference stage for a pre-trained network. The method exploits input data spatial locality via skipping partial processing of the multiply-and-accumulate (MAC) operations for adjacent data and equating its value to previously computed ones. SLID improves the throughput of resource-constrained devices (Internet-of-Things, edge devices) and accelerates computations during the inference phase by reducing the number of MAC operations. Such approximate computing schema does not require a similarity quantification step nor any modification for the training stage. The computational data reuse was evaluated on three well-known distinctive CNN structures and data sets with alternating layer selections: LeNet, CIFAR-10, and AlexNet. The computational data reuse method saves up to 34.9%, 49.84%, and 31.5% of MAC operations while reducing the accuracy by 8%, 3.7%, and 5.0% for the three models mentioned earlier, respectively. Besides, the proposed method saves on memory access by eliminating data fetching of skipped inputs. Furthermore, filter size, strides, and padding on the accuracy and savings of operations are analyzed. SLID is the first work to exploit the input spatial locality for savings on CNN convolution operations with minimal accuracy loss and without memory or computational overhead. This makes it a great option to support intelligence at the edge.

Natural language processing-based characterization of top-down communication in smart cities for enhancing citizen alignment

Many city governments have implemented promising smart initiatives to make cities more efficient, livable, and ecological. To harness the full potential of smart city initiatives, it is vital for policymakers to align citizens with the project objectives. This study comprehensively characterizes and classifies top-down announcements formulated by city developers into six alignment categories (i.e., smart economy, smart people, smart governance, smart mobility, smart environment, and smart living) using natural language processing. The proposed framework consists of five main processes: (1) web scraping-based extraction of announcements of four smart cities – Boston, Helsinki, Seoul, and Taipei, (2) text data preprocessing, (3) latent Dirichlet allocation-based modeling of strategic topics, (4) quantification of inter-topic similarities using Hellinger distance, and (5) comparison of top-down communication trends with real-world levels of urban performance. Through the comparative analysis of top-down communication trends and actual urban performances, the top-down discourses of smart cities were deconstructed as a reflection of wider political programs developed to enhance citizen alignment and urban performances. Furthermore, inter-topic similarities were also quantified to reflect whether communication strategies are multidisciplinary and city-tailored. In conclusion, the findings of this study can enhance our understanding and provide workable guidance for future smart city development.

Proposal for a Dynamic Composite Indicator: Application in a Comparative Analysis of Trends in the EU Member States Towards the Europe 2020 Strategy

This paper contributes, with a dynamic approach, to the research on the creation of comparable composite indicators by presenting a proposal for an exploratory factor analysis protocol to enable a comparative trend analysis. The originality of the study lies in the three dimensions of information for analysis: observations, variables and units of time. The proposal involves various stages of analysis with the ultimate, albeit not exclusive, aim of obtaining what is known as a Global Dynamic Indicator. The analysis process begins by structuring the data into a three-dimensional global matrix, thereby conditioning, while also, and primarily, enriching the later stages. A combination of multiple factor analysis and a clustering technique is the selected approach for successfully meeting the challenges involved. The appropriateness and versatility of the proposal are validated through the analysis of the trends of the EU member states towards the targets set by the 2020 Strategy. The study period runs from 2009 to 2018. The empirical work enables the visualisation and quantification of trend differences and similarities across member states collectively and individually, and across all the variables and years selected for analysis. The relevant findings will be quantified by means of a synthetic indicator for each unit of time and a global indicator for the period as a whole. Some of the conclusions reached by this paper are consistent with those already published by various authors.

Extraction of 3D quantitative maps using EDS-STEM tomography and HAADF-EDS bimodal tomography

Electron tomography has been widely applied to three-dimensional (3D) morphology characterization and chemical analysis at the nanoscale. A HAADF-EDS bimodal tomographic (HEBT) reconstruction technique has been developed to extract high resolution element-specific information. However, the reconstructed elemental maps cannot be directly converted to quantitative compositional information. In this work, we propose a quantification approach for obtaining elemental weight fraction maps from the HEBT reconstruction technique using the physical parameters extracted from a Monte Carlo code, MC X-ray. A similar quantification approach is proposed for the EDS-STEM tomographic reconstruction. The performance of the two quantitative reconstruction methods, using the simultaneous iterative reconstruction technique, are evaluated and compared for a simulated dataset of a two-dimensional phantom sample. The effects of the reconstruction parameters including the number of iterations and the weight of the HAADF signal are discussed. Finally, the two approaches are applied to an experimental dataset to show the 3D structure and quantitative elemental maps of a particle of flux melted metal-organic framework glass.

Diffusive gradients in thin films (DGT): A suitable tool for metals/metalloids monitoring in continental waterbodies at the large network scale

The contribution of Diffusive Gradients in Thin films (DGT) passive sampling to continental water quality monitoring was assessed in a real measurement network (6 sampling campaigns, 17 stations). Ten metals/metalloids (Al, Zn, Ni, Cd, Cu, Pb, Cr, As, Se and Sb) were studied using the control laboratory's working conditions with grab and DGT passive sampling. The DGT field deployments were robust, with a 3% sampler loss rate and a <65% average relative deviation between duplicates. Compared to grab sampling, DGT showed a similar quantification frequency for half of the targeted elements but showed a higher frequency for the other half (e.g., Cd quantification at 20% with grab sampling vs. 97% with DGT). Similar concentration trends were established using DGT and grab sampling at most sites throughout the year. Notably, for some elements, trends were only provided by DGT sampling. A study of several DGT blanks showed that the device contamination was occasional and originated primarily from cross-contamination during the disassembly step. Considering this contamination, the operational sensitivity by DGT was at least between 1 and 5 times greater in comparison to that by grab sampling.Estimations of the economic cost revealed that measurement networks cost 2 to 3 times more when monitored by DGT compared to standard grab monitoring. However, the information obtained based on each type of sampling method is different. Grab sampling is easy to implement and can highlight high contamination peaks. The DGT concentrations are averaged over time and are relevant to chronic exposure evaluations. Considering the good performance of the DGT sampling highlighted in this study and its complementarity with grab sampling in terms of water quality assessments, a combination of these two types of sampling, which can be affordable, should improve the water quality evaluation within monitoring networks.

Measuring Site-specific Glycosylation Similarity between Influenza a Virus Variants with Statistical Certainty

Influenza A virus (IAV) mutates rapidly, resulting in antigenic drift and poor year-to-year vaccine effectiveness. One challenge in designing effective vaccines is that genetic mutations frequently cause amino acid variations in IAV envelope protein hemagglutinin (HA) that create new N-glycosylation sequons; resulting N-glycans cause antigenic shielding, allowing viral escape from adaptive immune responses. Vaccine candidate strain selection currently involves correlating antigenicity with HA protein sequence among circulating strains, but quantitative comparison of site-specific glycosylation information may likely improve the ability to design vaccines with broader effectiveness against evolving strains. However, there is poor understanding of the influence of glycosylation on immunodominance, antigenicity, and immunogenicity of HA, and there are no well-tested methods for comparing glycosylation similarity among virus samples. Here, we present a method for statistically rigorous quantification of similarity between two related virus strains that considers the presence and abundance of glycopeptide glycoforms. We demonstrate the strength of our approach by determining that there was a quantifiable difference in glycosylation at the protein level between WT IAV HA from A/Switzerland/9715293/2013 (SWZ13) and a mutant strain of SWZ13, even though no N-glycosylation sequons were changed. We determined site-specifically that WT and mutant HA have varying similarity at the glycosylation sites of the head domain, reflecting competing pressures to evade host immune response while retaining viral fitness. To our knowledge, our results are the first to quantify changes in glycosylation state that occur in related proteins of considerable glycan heterogeneity. Our results provide a method for understanding how changes in glycosylation state are correlated with variations in protein sequence, which is necessary for improving IAV vaccine strain selection. Understanding glycosylation will be especially important as we find new expression vectors for vaccine production, as glycosylation state depends greatly on the host species.

Multiplex reverse transcriptase droplet digital PCR for the simultaneous quantification of four dengue serotypes: Proof of concept study

During vaccine production, RNA from chimeric yellow fever-dengue (CYD) vaccine viruses (CYD1, CYD2, CYD3 and CYD4) is currently quantified using separate serotype-specific RT-qPCR assays. Here we describe the results from a proof-of-concept study on the development of a multiplex reverse transcriptase droplet digital PCR (RT-ddPCR) assay for simultaneous quantification of RNA for all four viruses. Serotype-specific simplex RT-ddPCRs were developed using the serotype-specific PCR systems (forward and reverse primers and FAM (fluorescent chromophores 6-carboxyfluorescein) and YY (Yakima Yellow)-labelled probes), used in the routine RT-qPCR. The PCR systems were specific and gave similar quantification results to those from the RT-qPCR assay. Linear regression analyses were used to select relative probe concentrations to obtain distinct clusters for each target RNA in a 2-D cluster plot in a multiplex RT-ddPCR assay. We showed the clusters were positioned as predicted in the model for each CYD RNA and were well separated. The multiplex RT-ddPCR gave similar quantification results to those obtained by the serotype-specific RT-qPCR assays for triplicate samples containing 7, 8 or 9 Log10 Geq/mL. In conclusion, these results demonstrate that it is possible to quantify RNA from four CYD serotypes with a multiplex RT-ddPCR assay in a single assay.

Optimizing Numerical Simulations of Colliding Galaxies. I. Fitness Functions and Optimization Algorithms

Abstract Gravitational n-body models can be used to simulate the dynamical evolution of colliding galaxies. Given observational data in the form of images of the galaxies, we seek to estimate the true values of various dynamical parameters through the careful application of optimization methods. However, optimizing these models can be quite difficult due to (1) their expensive run-time, (2) the dimensionality and complexity of the parameter space which must be explored, (3) the presence of morphological symmetries which introduce many false extrema into the fitness landscape, and (4) the fact that the majority of orbital trajectories produce systems with little to no tidal features. To combat these issues, we have developed novel “two-factor” fitness functions for image similarity quantification and an adaptive genetic algorithm with crossover and mutation operators designed to handle parameter correlation. We test these methods by fitting to a synthetic target image.

Rapid Quantification of White Matter Disconnection in the Human Brain.

With an estimated five million new stroke survivors every year and a rapidly aging population suffering from hyperintensities and diseases of presumed vascular origin that affect white matter and contribute to cognitive decline, it is critical that we understand the impact of white matter damage on brain structure and behavior. Current techniques for assessing the impact of lesions consider only location, type, and extent, while ignoring how the affected region was connected to the rest of the brain. Regional brain function is a product of both local structure and its connectivity. Therefore, obtaining a map of white matter disconnection is a crucial step that could help us predict the behavioral deficits that patients exhibit. In the present work, we introduce a new practical method for computing lesion-based white matter disconnection maps that require only moderate computational resources. We achieve this by creating diffusion tractography models of the brains of healthy adults and assessing the connectivity between small regions. We then interrupt these connectivity models by projecting patients' lesions into them to compute predicted white matter disconnection. A quantified disconnection map can be computed for an individual patient in approximately 35 seconds using a single core CPU-based computation. In comparison, a similar quantification performed with other tools provided by MRtrix3 takes 5.47 minutes.

Application of propidium monoazide coupled with quantitative PCR to evaluate cell viability of Bifidobacterium animalis subsp. lactis in a non-dairy probiotic beverage

PurposeIn this study, a PMA-qPCR assay was developed for the enumeration of Bifidobacterium animalis subsp. lactis BB-12 viable cells in a non-dairy probiotic beverage.MethodsProbiotic viability was monitored in three formulations of probiotic passion fruit juice microencapsulated by spray drying, during 30 days of storage at 4 °C. Viable cells were quantified using qPCR and PMA-qPCR assays targeting tuf gene and by plate counting method.ResultsThe limit of detection for all samples was 103 genome copies, corresponding to 21.3 pg of DNA. Higher CFU values were obtained for B. lactis BB-12 enumeration by qPCR, when compared to those obtained by PMA-qPCR and plate count, for all probiotic juice microcapsules. Similar quantification values were obtained by PMA-qPCR and plate counting for all samples and remained above 8 log CFU/g during the storage period.ConclusionThese results demonstrated that the PMA-qPCR technique is a promising approach for B. lactis BB-12 viable cell enumeration in complex matrices such as passion fruit juice microcapsules. This PMA-qPCR assay allowed the achievement of reliable results faster than with the traditional plate counting method.

Local structure order parameters and site fingerprints for quantification of coordination environment and crystal structure similarity.

Structure characterization and classification is frequently based on local environment information of all or selected atomic sites in the crystal structure. Therefore, reliable and robust procedures to find coordinated neighbors and to evaluate the resulting coordination pattern (e.g., tetrahedral, square planar) are critically important for both traditional and machine learning approaches that aim to exploit site or structure information for predicting materials properties. Here, we introduce new local structure order parameters (LoStOPs) that are specifically designed to rapidly detect highly symmetric local coordination environments (e.g., Platonic solids such as a tetrahedron or an octahedron) as well as less symmetric ones (e.g., Johnson solids such as a square pyramid). Furthermore, we introduce a Monte Carlo optimization approach to ensure that the different LoStOPs are comparable with each other. We then apply the new local environment descriptors to define site and structure fingerprints and to measure similarity between 61 known coordination environments and 40 commonly studied crystal structures, respectively. After extensive testing and optimization, we determine the most accurate structure similarity assessment procedure to compute all 2.45 billion structure similarities between each pair of the ≈70 000 materials that are currently present in the Materials Project database.

A tale of two niches: methods, concepts, and evolution

Being snapshots in time, species ranges may fall short of representing all of the geographic or environmental space that they are able to occupy. This has important implications for niche studies yet most comparative studies overlook the transient nature of species distributions and assume that they are at equilibrium. We review the methods most widely used for niche comparisons today and suggest a modified framework to describe and compare niches based on snapshot species range data. First, we introduce a new environmental space-based Niche Equivalence Statistic to test niche similarity between two species, which explicitly incorporates the spatial distribution of environments and their availability into statistical tests. We also introduce a new Background Statistic to measure the ability of this Niche Equivalence Statistic to detect differences based on the available environmental-space. These metrics enable fair comparisons between different geographies when the ranges of species are out of equilibrium. Based on distinct parameterizations of the new Equivalence and Background statistics, we then propose a Niche Divergence Test and a Niche Overlap Test, which allow assessment of whether differences between species emerge from true niche divergences. These methods are implemented in a new R package, ‘humboldt’ and applied to simulated species with pre-defined niches. The new methods improve accuracy of niche similarity and associated tests – consistently outperforming other tests. We show that the quantification of niche similarity should be performed only in environmental space, which is less sensitive than geographic space to the spatial abundance of key environmental variables. Further, our methods characterize the relationships between non-analogous and analogous climates in the species’ distributions, something not available previously. These improvements allow assessment of whether the different environmental spaces occupied by two taxa emerge from true niche evolution, as opposed to differences in life history and biological interactors, or differences in the variety and configuration of environments accessible to them.

Assessment of differences between DNA content of cell-cultured and freely suspended oocysts of Cryptosporidium parvum and their suitability as DNA standards in qPCR

BackgroundAlthough more modern methods are available, quantitative PCR (qPCR) is reproducible, sensitive and specific with instruments and expertise readily available in many laboratories. As such, the use of qPCR in Cryptosporidium research is well established and still widely used by researchers globally. This method depends upon the generation of standards at different concentrations to generate standard curves subsequently used for the quantification of DNA.MethodsWe assessed four types of DNA template used to generate standard curves in drug screening studies involving Cryptosporidium spp.: (i) serially diluted Cryptosporidium parvum oocysts (106–1); (ii) diluted template DNA from pure oocysts (×10–×106 dilution of 106 oocyst DNA template); (iii) oocysts incubated in human ileocecal adenocarcinoma (HCT-8) cells (105–1 and 5 × 104–50); and (iv) diluted DNA template (5 × 104) from cell culture incubated parasites (×10–×1000).ResultsSerial dilutions of both cell culture and pure oocyst suspension DNA template yielded better linearity than cell culture derived standards, with dilutions of 106 oocysts exhibiting similar quantification cycle (Cq) values to those obtained from DNA template dilutions of 106 oocysts. In contrast, cell culture incubated oocysts demonstrated significantly higher DNA content than equivalent freely suspended oocysts and diluted DNA template from both cell culture derived and freely suspended oocysts across numerous concentrations.ConclusionsFor many studies involving Cryptosporidium, only relative DNA content is required and as such, the superior linearity afforded by freely suspended oocysts and diluted DNA template (from either cell culture derived standards or freely suspended oocysts) will allow for more accurate relative quantification in each assay. Parasite division in the cell culture standards likely explains the higher DNA content found. These standards, therefore, have the potential to more accurately reflect DNA content in cell culture assays, and despite more modern methods available for absolute quantification, i.e. droplet digital PCR (ddPCR), the ubiquity of qPCR for the foreseeable future encourages further investigation into the reduced linearity observed in these standards such as varying oocyst seeding density, non-linear growth rates and assay efficiency.

Image acquisition stability of fixated musculoskeletal sonography in an exercise setting: a quantitative analysis and comparison with freehand acquisition.

In dynamic musculoskeletal sonography, probe fixation can contribute to field of view (FOV) consistency, which is necessary for valid analysis of architectural parameters. In this volunteer study, the achieved FOV consistency in fixated ultrasonography was quantified and compared with freehand acquisition. During five resting periods during cycling exercise, longitudinal B-mode images of the vastus lateralis (VL) muscle were acquired on one thigh with a fixated probe, and by two trained observers on the other thigh. In each acquisition, the structural similarity compared to the first resting period was determined using the complex wavelet structural similarity index (CW-SSIM). Also, the pennation angle of the VL was measured. Both CW-SSIM and pennation angle were compared between fixated and freehand acquisition. Furthermore, the compression of tissue by the probe fixation was measured. In fixated acquisition, a significantly higher structural similarity (p < 0.05) and an improved repeatability of pennation angle measurement were obtained compared to freehand acquisition. Probe fixation compressed muscle tissue by 12% on average. Quantification of the structural similarity showed an increase in FOV consistency with sonography compared to freehand acquisition. The demonstrated feasibility of long-term fixated acquisition might be attractive in many medical fields and sports, and for reduction of work-related ergonomic problems among sonographers.

Quantifying semantic similarity of clinical evidence in the biomedical literature to facilitate related evidence synthesis.

Published clinical trials and high quality peer reviewed medical publications are considered as the main sources of evidence used for synthesizing systematic reviews or practicing Evidence Based Medicine (EBM). Finding all relevant published evidence for a particular medical case is a time and labour intensive task, given the breadth of the biomedical literature. Automatic quantification of conceptual relationships between key clinical evidence within and across publications, despite variations in the expression of clinically-relevant concepts, can help to facilitate synthesis of evidence. In this study, we aim to provide an approach towards expediting evidence synthesis by quantifying semantic similarity of key evidence as expressed in the form of individual sentences. Such semantic textual similarity can be applied as a key approach for supporting selection of related studies. We propose a generalisable approach for quantifying semantic similarity of clinical evidence in the biomedical literature, specifically considering the similarity of sentences corresponding to a given type of evidence, such as clinical interventions, population information, clinical findings, etc. We develop three sets of generic, ontology-based, and vector-space models of similarity measures that make use of a variety of lexical, conceptual, and contextual information to quantify the similarity of full sentences containing clinical evidence. To understand the impact of different similarity measures on the overall evidence semantic similarity quantification, we provide a comparative analysis of these measures when used as input to an unsupervised linear interpolation and a supervised regression ensemble. In order to provide a reliable test-bed for this experiment, we generate a dataset of 1000 pairs of sentences from biomedical publications that are annotated by ten human experts. We also extend the experiments on an external dataset for further generalisability testing. The combination of all diverse similarity measures showed stronger correlations with the gold standard similarity scores in the dataset than any individual kind of measure. Our approach reached near 0.80 average Pearson correlation across different clinical evidence types using the devised similarity measures. Although they were more effective when combined together, individual generic and vector-space measures also resulted in strong similarity quantification when used in both unsupervised and supervised models. On the external dataset, our similarity measures were highly competitive with the state-of-the-art approaches developed and trained specifically on that dataset for predicting semantic similarity. Experimental results showed that the proposed semantic similarity quantification approach can effectively identify related clinical evidence that is reported in the literature. The comparison with a state-of-the-art method demonstrated the effectiveness of the approach, and experiments with an external dataset support its generalisability.

Enactive experience promotes early number understanding: a study with 3-year-old children

ABSTRACTWe investigated the hypothesis that embodied action improves early numerical cognition and learning by means of an experimental study with young children. The influence of children’s enactive participation was analysed in a playful situation that posed different number-related tasks regarding quantities 1–5. Participants were 44 3-year-old children who were randomly assigned to one of two conditions: Enaction or Observation. In the Enaction group, children were encouraged to establish a direct contact with the playful materials, while in the Observation group they were encouraged to observe the demonstrations by the experimenter. Participants in both conditions started from a similar baseline quantification performance; however, the Enaction group systematically outperformed the Observation group across number tasks. This was so both when overall performance was compared and when fine-grained analyses by magnitude within tasks were conducted. These robust findings point at the beneficial impact of enactive experience for initial number understanding.

Quality evaluation of bee pollens by chromatographic fingerprint and simultaneous determination of its major bioactive components

Bee pollens constitute a large number of flavonoids and thus possess great medicinal value. However different varieties of bee pollen flavonoids vary with different species and their content also differ greatly in different region. Herein, the aim of present research is to establish a method based on high performance liquid chromatography (HPLC) for quantitative analysis of flavonoids compounds and chemical fingerprint analysis of bee pollen. Five batches of rape bee pollen collected from different region of China and particularly six bee pollen species obtained in Anhui were used to establish the fingerprint. The feasibility and advantages of the used HPLC fingerprint were verified for its similarity evaluation by systematically comparing chromatograms with professional analytical software. The similarities of liquid chromatography fingerprints for five batches of rape bee pollen were more than 0.994 while six batches of different species of bee pollen were lower than 0.810. In quantitative analysis, the six compounds showed good regression (R ≥ 0.9964) within the test ranges, and all the values for the RSD were lower than 2%. The developed HPLC fingerprint method was found simple, reliable, and it was validated for the quality control and identification of bee pollen. Additionally, simultaneous quantification of six flavonoids ingredients in the bee pollen samples was conducted to reveal the variation in their content. The results indicated that the HPLC fingerprint, as a characteristic distinguishing method combining similarity evaluation and quantification analysis, can be successfully used to assess the quality and also to identify the authenticity of bee pollen.

Topology of Networks in Generalized Musical Spaces

The abstraction of musical structures as mathematical objects in a geometrical space is one of the major accomplishments of contemporary music theory. The author generalizes the concept of musical spaces as networks and derives compositional design principles via network topology analysis. This approach provides a framework for analysis and quantification of similarity of musical objects and structures and suggests a way to relate such measures to human perception of different musical entities. Finally, network analysis provides alternative ways of interpreting the compositional process by quantifying emergent behaviors with well-established statistical mechanics. Interpreting the latter as probabilistic randomness in the network, the author develops novel compositional design frameworks.

A Concept for Graph-Based Temporal Similarity of Patient Data.

Computer-based decision support systems are often used for dedicated tasks such as the detection of sepsis. However, positive predictive values for sepsis detection are reported to achieve only around 46%. In this paper we describe a novel approach to use temporal data of electronic patient records based on similarity measures. We apply the concept of case-based reasoning, which is well-established in many fields of medical informatics. Temporal patient data are organized in a time-graph structure. For the quantification of similarity between cases, we exploit graph theory based approaches. For development and evaluation of our time-graph similarity frame we use the open MIMIC III dataset. In a later phase, we envision to transfer our concept from sepsis to other diseases.