Krzysztof Cios - Academia.edu (original) (raw)
Papers by Krzysztof Cios
<p>Figures illustrating each SOM and the classes compared in each are indicated. Where more... more <p>Figures illustrating each SOM and the classes compared in each are indicated. Where more than two classes were compared (columns 1, 4, 7 and 8), only those proteins common to all comparisons are listed; complete lists of proteins for each comparison are provided in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s003" target="_blank">S1</a> and <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s004" target="_blank">S2</a> Tables. MAPK, components of the MAP kinase pathway; MTOR, components of the mechanistic target of rapamycin pathway; AD, proteins observed to be abnormal in brains from patients with or mouse models of Alzheimer’s Disease; NMDAR, subunits of ionotropic glutamate receptors and interacting proteins; Hsa21, proteins encoded by human chromosomes 21; IEG, immediate early gene proteins; apoptosis-related, BAD, proapoptotic, BCL2, antiapoptotic; histone, histone protein H3 modifications: Ac, acetylation, Me, methylation, K amino acid number of modified lysine residue; Misc, miscellaneous.</p><p>Functional associations of discriminant proteins used to generate SOMs.</p
IEEE Transactions on Biomedical Engineering, 2011
Multisensory processing in the brain underlies a wide variety of perceptual phenomena, but little... more Multisensory processing in the brain underlies a wide variety of perceptual phenomena, but little is known about the underlying mechanisms of how multisensory neurons are formed. This lack of knowledge is due to the difficulty for biological experiments to manipulate and test the parameters of multisensory convergence, the first and definitive step in the multisensory process. Therefore, by using a computational model of multisensory convergence, this study seeks to provide insight into the mechanisms of multisensory convergence. To reverse-engineer multisensory convergence, we used a biologically realistic neuron model and a biology-inspired plasticity rule, but did not make any a priori assumptions about multisensory properties of neurons in the network. The network consisted of two separate projection areas that converged upon neurons in a third area, and stimulation involved activation of one of the projection areas (or the other) or their combination. Experiments consisted of two parts: network training and multisensory simulation. Analyses were performed, first, to find multisensory properties in the simulated networks; second, to reveal properties of the network using graph theoretical approach; and third, to generate hypothesis related to the multisensory convergence. The results showed that the generation of multisensory neurons related to the topological properties of the network, in particular, the strengths of connections after training, was found to play an important role in forming and thus distinguishing multisensory neuron types. Index Terms-Computational modeling, multisensory convergence, network of spiking neurons, reverse engineering. I. INTRODUCTION O NE of the challenges in neuroscience is to understand the factors that govern multisensory processing because this phenomenon underlies a wide variety of perceptual phenomena. Analogously, one of the challenges in engineering is to design smarter artificial systems that would perform difficult (for computers) perceptual tasks, such as image recognition, at the level Manuscript
IEEE Engineering in Medicine and Biology Magazine, 2007
Abstract The article presents a new, comprehensive, user-friendly chromosome 21 database driven b... more Abstract The article presents a new, comprehensive, user-friendly chromosome 21 database driven by a built-in protein interaction prediction tool based on Markov random field (MRF) and GeneQuest, a novel easy-to-use user interface. The database contains a wide range ...
IEEE Engineering in Medicine and Biology Magazine, 1994
ABSTRACT
IEEE Engineering in Medicine and Biology Magazine, 2005
ABSTRACT The paper aims to develop an automated system that would ensure a robust peptide quantif... more ABSTRACT The paper aims to develop an automated system that would ensure a robust peptide quantification process, which would permit researchers to quantify desired proteins faster and with greater reliability. Because of the uniqueness of the data used, biochemists' expertise and data mining methods were employed in this work. The system includes two main system components: one for the discovery of two quantification peptides and the internal standard peptide and the other for protein quantification in patient samples. If the required input data are available, each subsystem can be run separately. The developed system can be applied to similar problems because our design is flexible, allowing for easy adaptation.
<p><b>(A)</b> Clustering of trisomic mice with proteins common to the two compa... more <p><b>(A)</b> Clustering of trisomic mice with proteins common to the two comparisons that reflected rescued learning: t-CS-m vs. t-SC-m and t-CS-m vs. t-SC-s (15 proteins), plus the initial effects of memantine: t-SC-m vs. t-SC-s (12 proteins). The t-SC-m cluster is outlined in solid blue and the t-SC-s cluster is outlined in dashed blue. <b>(B)</b> Clustering with the former 15 proteins (rescued learning) plus the 9 discriminant between t-CS-m and t-CS-s (<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s004" target="_blank">S2 Table</a>, c5).</p
[1992] Proceedings of the IEEE International Symposium on Industrial Electronics
ABSTRACT
Applied Soft Computing, 2015
One-class learning algorithms are used in situations when training data are available only for on... more One-class learning algorithms are used in situations when training data are available only for one class, called target class. Data for other class(es), called outliers, are not available. One-class learning algorithms are used for detecting outliers, or novelty, in the data. The common approachin one-class learning is to use density estimation techniques or adapt standard classification algorithms to define a decision boundary that encompasses only the target data. In this paper, we introduce OneClass-DS learning algorithm that combines rule-based classification with greedy search algorithm based on density of features. Its performance is tested on 25 data sets and compared with eight other one-class algorithms; the results show that it performs on par with thosealgorithms.
2013 6th International IEEE/EMBS Conference on Neural Engineering (NER), 2013
Izhikevich model of a neuron allows for simulation of spiking pattern that mimics known biologica... more Izhikevich model of a neuron allows for simulation of spiking pattern that mimics known biological subtypes. When a current within a range typical for biological experiments is injected into the cell the firing pattern produced in the simulation is close to that observed biologically. However, once these neurons are embedded into a network, the level of depolarization is controlled only by the synaptic depolarization received by the simulated connections. Under these conditions there is no limit on the maximum firing rate produced by any of the neurons. Here we introduce a modification of the Izhikevich model to restrict the firing rate. We demonstrate how this modification affects the overall network activity using a simple artificial neural network. The proposed restraint on the Izhikevich model is particularly important for larger scale simulations or when the frequency dependent short-term plasticity is used in the network. Although maximum firing rates are most likely exceeded in simulations of seizure-like activity we show that restriction of neuronal firing frequencies impacts even small networks with moderate levels of activity.
PLOS ONE, 2015
Down syndrome (DS) is a chromosomal abnormality (trisomy of human chromosome 21) associated with ... more Down syndrome (DS) is a chromosomal abnormality (trisomy of human chromosome 21) associated with intellectual disability and affecting approximately one in 1000 live births worldwide. The overexpression of genes encoded by the extra copy of a normal chromosome in DS is believed to be sufficient to perturb normal pathways and normal responses to stimulation, causing learning and memory deficits. In this work, we have designed a strategy based on the unsupervised clustering method, Self Organizing Maps (SOM), to identify biologically important differences in protein levels in mice exposed to context fear conditioning (CFC). We analyzed expression levels of 77 proteins obtained from normal genotype control mice and from their trisomic littermates (Ts65Dn) both with and without treatment with the drug memantine. Control mice learn successfully while the trisomic mice fail, unless they are first treated with the drug, which rescues their learning ability. The SOM approach identified reduced subsets of proteins predicted to make the most critical contributions to normal learning, to failed learning and rescued learning, and provides a visual representation of the data that allows the user to extract patterns that may underlie novel biological responses to the different kinds of learning and the response to memantine. Results suggest that the application of SOM to new experimental data sets of complex protein profiles can be used to identify common critical protein responses, which in turn may aid in identifying potentially more effective drug targets.
Studies in Fuzziness and Soft Computing, 2002
Data Mining
This Chapter provides background and algorithms for feature extraction and feature selection from... more This Chapter provides background and algorithms for feature extraction and feature selection from numerical data. Both methods are performed to reduce the dimensionality of the original data. Feature extraction methods do it by generating new transformed features and selecting the informative ones while feature selection methods choose a subset of original features.
Studies in Computational Intelligence, 2010
Handbook of Neural Computation, 1996
Handbook of Neural Computation, 1996
Current Bioinformatics, 2008
Recent years observed a growing interest in computational methods that predict and characterize p... more Recent years observed a growing interest in computational methods that predict and characterize protein structure due to the increasing sequence-structure gap. This includes a spike in development of sequence-based in-silico methods that address prediction of several newly formulated real-value descriptors of protein structure. These descriptors include B-factor, backbone torsion angles, solvent accessibility, residue depth, contact number, residue-wise contact order, secondary structure content, and folding rates. Although they address different structural aspects, such as exposure to the solvent, spatial position and packing of the residues, their flexibility, amount of secondary structures in the protein, and folding time, the methods that are built to address them share similarities that could be exploited to improve future designs. To date, no comprehensive overview that summarizes and contrasts solutions developed for these tasks was published. To address this we compare different designs of real-value predictors based on information concerning input data encoding and prediction algorithms used. We also investigate evaluation standards, which include benchmark datasets, test criteria, and test procedures used in these predictive tasks. Finally, we summarize application areas and problems that use the above-mentioned predictions. We believe that the breath and number of these applications justify further development of more accurate and integrated real-value prediction methods.
<p>Figures illustrating each SOM and the classes compared in each are indicated. Where more... more <p>Figures illustrating each SOM and the classes compared in each are indicated. Where more than two classes were compared (columns 1, 4, 7 and 8), only those proteins common to all comparisons are listed; complete lists of proteins for each comparison are provided in <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s003" target="_blank">S1</a> and <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s004" target="_blank">S2</a> Tables. MAPK, components of the MAP kinase pathway; MTOR, components of the mechanistic target of rapamycin pathway; AD, proteins observed to be abnormal in brains from patients with or mouse models of Alzheimer’s Disease; NMDAR, subunits of ionotropic glutamate receptors and interacting proteins; Hsa21, proteins encoded by human chromosomes 21; IEG, immediate early gene proteins; apoptosis-related, BAD, proapoptotic, BCL2, antiapoptotic; histone, histone protein H3 modifications: Ac, acetylation, Me, methylation, K amino acid number of modified lysine residue; Misc, miscellaneous.</p><p>Functional associations of discriminant proteins used to generate SOMs.</p
IEEE Transactions on Biomedical Engineering, 2011
Multisensory processing in the brain underlies a wide variety of perceptual phenomena, but little... more Multisensory processing in the brain underlies a wide variety of perceptual phenomena, but little is known about the underlying mechanisms of how multisensory neurons are formed. This lack of knowledge is due to the difficulty for biological experiments to manipulate and test the parameters of multisensory convergence, the first and definitive step in the multisensory process. Therefore, by using a computational model of multisensory convergence, this study seeks to provide insight into the mechanisms of multisensory convergence. To reverse-engineer multisensory convergence, we used a biologically realistic neuron model and a biology-inspired plasticity rule, but did not make any a priori assumptions about multisensory properties of neurons in the network. The network consisted of two separate projection areas that converged upon neurons in a third area, and stimulation involved activation of one of the projection areas (or the other) or their combination. Experiments consisted of two parts: network training and multisensory simulation. Analyses were performed, first, to find multisensory properties in the simulated networks; second, to reveal properties of the network using graph theoretical approach; and third, to generate hypothesis related to the multisensory convergence. The results showed that the generation of multisensory neurons related to the topological properties of the network, in particular, the strengths of connections after training, was found to play an important role in forming and thus distinguishing multisensory neuron types. Index Terms-Computational modeling, multisensory convergence, network of spiking neurons, reverse engineering. I. INTRODUCTION O NE of the challenges in neuroscience is to understand the factors that govern multisensory processing because this phenomenon underlies a wide variety of perceptual phenomena. Analogously, one of the challenges in engineering is to design smarter artificial systems that would perform difficult (for computers) perceptual tasks, such as image recognition, at the level Manuscript
IEEE Engineering in Medicine and Biology Magazine, 2007
Abstract The article presents a new, comprehensive, user-friendly chromosome 21 database driven b... more Abstract The article presents a new, comprehensive, user-friendly chromosome 21 database driven by a built-in protein interaction prediction tool based on Markov random field (MRF) and GeneQuest, a novel easy-to-use user interface. The database contains a wide range ...
IEEE Engineering in Medicine and Biology Magazine, 1994
ABSTRACT
IEEE Engineering in Medicine and Biology Magazine, 2005
ABSTRACT The paper aims to develop an automated system that would ensure a robust peptide quantif... more ABSTRACT The paper aims to develop an automated system that would ensure a robust peptide quantification process, which would permit researchers to quantify desired proteins faster and with greater reliability. Because of the uniqueness of the data used, biochemists' expertise and data mining methods were employed in this work. The system includes two main system components: one for the discovery of two quantification peptides and the internal standard peptide and the other for protein quantification in patient samples. If the required input data are available, each subsystem can be run separately. The developed system can be applied to similar problems because our design is flexible, allowing for easy adaptation.
<p><b>(A)</b> Clustering of trisomic mice with proteins common to the two compa... more <p><b>(A)</b> Clustering of trisomic mice with proteins common to the two comparisons that reflected rescued learning: t-CS-m vs. t-SC-m and t-CS-m vs. t-SC-s (15 proteins), plus the initial effects of memantine: t-SC-m vs. t-SC-s (12 proteins). The t-SC-m cluster is outlined in solid blue and the t-SC-s cluster is outlined in dashed blue. <b>(B)</b> Clustering with the former 15 proteins (rescued learning) plus the 9 discriminant between t-CS-m and t-CS-s (<a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0129126#pone.0129126.s004" target="_blank">S2 Table</a>, c5).</p
[1992] Proceedings of the IEEE International Symposium on Industrial Electronics
ABSTRACT
Applied Soft Computing, 2015
One-class learning algorithms are used in situations when training data are available only for on... more One-class learning algorithms are used in situations when training data are available only for one class, called target class. Data for other class(es), called outliers, are not available. One-class learning algorithms are used for detecting outliers, or novelty, in the data. The common approachin one-class learning is to use density estimation techniques or adapt standard classification algorithms to define a decision boundary that encompasses only the target data. In this paper, we introduce OneClass-DS learning algorithm that combines rule-based classification with greedy search algorithm based on density of features. Its performance is tested on 25 data sets and compared with eight other one-class algorithms; the results show that it performs on par with thosealgorithms.
2013 6th International IEEE/EMBS Conference on Neural Engineering (NER), 2013
Izhikevich model of a neuron allows for simulation of spiking pattern that mimics known biologica... more Izhikevich model of a neuron allows for simulation of spiking pattern that mimics known biological subtypes. When a current within a range typical for biological experiments is injected into the cell the firing pattern produced in the simulation is close to that observed biologically. However, once these neurons are embedded into a network, the level of depolarization is controlled only by the synaptic depolarization received by the simulated connections. Under these conditions there is no limit on the maximum firing rate produced by any of the neurons. Here we introduce a modification of the Izhikevich model to restrict the firing rate. We demonstrate how this modification affects the overall network activity using a simple artificial neural network. The proposed restraint on the Izhikevich model is particularly important for larger scale simulations or when the frequency dependent short-term plasticity is used in the network. Although maximum firing rates are most likely exceeded in simulations of seizure-like activity we show that restriction of neuronal firing frequencies impacts even small networks with moderate levels of activity.
PLOS ONE, 2015
Down syndrome (DS) is a chromosomal abnormality (trisomy of human chromosome 21) associated with ... more Down syndrome (DS) is a chromosomal abnormality (trisomy of human chromosome 21) associated with intellectual disability and affecting approximately one in 1000 live births worldwide. The overexpression of genes encoded by the extra copy of a normal chromosome in DS is believed to be sufficient to perturb normal pathways and normal responses to stimulation, causing learning and memory deficits. In this work, we have designed a strategy based on the unsupervised clustering method, Self Organizing Maps (SOM), to identify biologically important differences in protein levels in mice exposed to context fear conditioning (CFC). We analyzed expression levels of 77 proteins obtained from normal genotype control mice and from their trisomic littermates (Ts65Dn) both with and without treatment with the drug memantine. Control mice learn successfully while the trisomic mice fail, unless they are first treated with the drug, which rescues their learning ability. The SOM approach identified reduced subsets of proteins predicted to make the most critical contributions to normal learning, to failed learning and rescued learning, and provides a visual representation of the data that allows the user to extract patterns that may underlie novel biological responses to the different kinds of learning and the response to memantine. Results suggest that the application of SOM to new experimental data sets of complex protein profiles can be used to identify common critical protein responses, which in turn may aid in identifying potentially more effective drug targets.
Studies in Fuzziness and Soft Computing, 2002
Data Mining
This Chapter provides background and algorithms for feature extraction and feature selection from... more This Chapter provides background and algorithms for feature extraction and feature selection from numerical data. Both methods are performed to reduce the dimensionality of the original data. Feature extraction methods do it by generating new transformed features and selecting the informative ones while feature selection methods choose a subset of original features.
Studies in Computational Intelligence, 2010
Handbook of Neural Computation, 1996
Handbook of Neural Computation, 1996
Current Bioinformatics, 2008
Recent years observed a growing interest in computational methods that predict and characterize p... more Recent years observed a growing interest in computational methods that predict and characterize protein structure due to the increasing sequence-structure gap. This includes a spike in development of sequence-based in-silico methods that address prediction of several newly formulated real-value descriptors of protein structure. These descriptors include B-factor, backbone torsion angles, solvent accessibility, residue depth, contact number, residue-wise contact order, secondary structure content, and folding rates. Although they address different structural aspects, such as exposure to the solvent, spatial position and packing of the residues, their flexibility, amount of secondary structures in the protein, and folding time, the methods that are built to address them share similarities that could be exploited to improve future designs. To date, no comprehensive overview that summarizes and contrasts solutions developed for these tasks was published. To address this we compare different designs of real-value predictors based on information concerning input data encoding and prediction algorithms used. We also investigate evaluation standards, which include benchmark datasets, test criteria, and test procedures used in these predictive tasks. Finally, we summarize application areas and problems that use the above-mentioned predictions. We believe that the breath and number of these applications justify further development of more accurate and integrated real-value prediction methods.