Evgeny M. Mirkes

A random six-phase switch regulates pneumococcal virulence via global epigenetic changes

Streptococcus pneumoniae (the pneumococcus) is the world’s foremost bacterial pathogen in both morbidity and mortality. Switching between phenotypic forms (or ‘phases’) that favour asymptomatic carriage or invasive disease was first reported in 1933. Here, we show that the underlying mechanism for such phase variation consists of genetic rearrangements in a Type I restriction-modification system (SpnD39III). The rearrangements generate six alternative specificities with distinct methylation patterns, as defined by single-molecule, real-time (SMRT) methylomics. The SpnD39III variants have distinct gene expression profiles. We demonstrate distinct virulence in experimental infection and in vivo selection for switching between SpnD39III variants. SpnD39III is ubiquitous in pneumococci, indicating an essential role in its biology. Future studies must recognize the potential for switching between these heretofore undetectable, differentiated pneumococcal subpopulations in vitro and in vivo. Similar systems exist in other bacterial genera, indicating the potential for broad exploitation of epigenetic gene regulation.

Data complexity measured by principal graphs

Abstract How to measure the complexity of a finite set of vectors embedded in a multidimensional space? This is a non-trivial question which can be approached in many different ways. Here we suggest a set of data complexity measures using universal approximators, principal cubic complexes. Principal cubic complexes generalize the notion of principal manifolds for datasets with non-trivial topologies. The type of the principal cubic complex is determined by its dimension and a grammar of elementary graph transformations. The simplest grammar produces principal trees. We introduce three natural types of data complexity: (1) geometric (deviation of the data’s approximator from some “idealized” configuration, such as deviation from harmonicity); (2) structural (how many elements of a principal graph are needed to approximate the data), and (3) construction complexity (how many applications of elementary graph transformations are needed to construct the principal object starting from the simplest one). We compute these measures for several simulated and real-life data distributions and show them in the “accuracy–complexity” plots, helping to optimize the accuracy/complexity ratio. We discuss various issues connected with measuring data complexity. Software for computing data complexity measures from principal cubic complexes is provided as well.

Learning optimization for decision tree classification of non-categorical data with information gain impurity criterion

We consider the problem of construction of decision trees in cases when data is non-categorical and is inherently high-dimensional. Using conventional tree growing algorithms that either rely on univariate splits or employ direct search methods for determining multivariate splitting conditions is computationally prohibitive. On the other hand application of standard optimization methods for finding locally optimal splitting conditions is obstructed by abundance of local minima and discontinuities of classical goodness functions such as e.g. information gain or Gini impurity. In order to avoid this limitation a method to generate smoothed replacement for measuring impurity of splits is proposed. This enables to use vast number of efficient optimization techniques for finding locally optimal splits and, at the same time, decreases the number of local minima. The approach is illustrated with examples.

Computational diagnosis of canine lymphoma

One out of four dogs will develop cancer in their lifetime and 20% of those will be lymphoma cases. PetScreen developed a lymphoma blood test using serum samples collected from several veterinary practices. The samples were fractionated and analysed by mass spectrometry. Two protein peaks, with the highest diagnostic power, were selected and further identified as acute phase proteins, C-Reactive Protein and Haptoglobin. Data mining methods were then applied to the collected data for the development of an online computer-assisted veterinary diagnostic tool. The generated software can be used as a diagnostic, monitoring and screening tool. Initially, the diagnosis of lymphoma was formulated as a classification problem and then later refined as a lymphoma risk estimation. Three methods, decision trees, kNN and probability density evaluation, were used for classification and risk estimation and several pre- processing approaches were implemented to create the diagnostic system. For the differential diagnosis the best solution gave a sensitivity and specificity of 83.5% and 77%, respectively (using three input features, CRP, Haptoglobin and standard clinical symptom). For the screening task, the decision tree method provided the best result, with sensitivity and specificity of 81.4% and >99%, respectively (using the same input features). Furthermore, the development and application of new techniques for the generation of risk maps allowed their user-friendly visualization.

Fluorescence-based assay as a new screening tool for toxic chemicals.

Our study involves development of fluorescent cell-based diagnostic assay as a new approach in high-throughput screening method. This highly sensitive optical assay operates similarly to e-noses and e-tongues which combine semi-specific sensors and multivariate data analysis for monitoring biochemical processes. The optical assay consists of a mixture of environmental-sensitive fluorescent dyes and human skin cells that generate fluorescence spectra patterns distinctive for particular physico-chemical and physiological conditions. Using chemometric techniques the optical signal is processed providing qualitative information about analytical characteristics of the samples. This integrated approach has been successfully applied (with sensitivity of 93% and specificity of 97%) in assessing whether particular chemical agents are irritating or not for human skin. It has several advantages compared with traditional biochemical or biological assays and can impact the new way of high-throughput screening and understanding cell activity. It also can provide reliable and reproducible method for assessing a risk of exposing people to different harmful substances, identification active compounds in toxicity screening and safety assessment of drugs, cosmetic or their specific ingredients.

Thermodynamic consistency of kinetic data

It is well known [1-4] that the rate constants of different elementary reactions are often interdependent. Relationships determined by the principle of detailed balancing exist between them [1-5] when microreversibility is valid and by the generalizations of that principle [5-8] when it is not (for example, in magnetic fields, during macroscopic rotations, etc.). Nevertheless, in practice the verification of consistency in the kinetic constants for complicated transformation schemes involves a certain amount of technical difficulty.

The Five Factor Model of Personality and Evaluation of Drug Consumption Risk

The problem of evaluating an individual’s risk of drug consumption and misuse is highly important and novel. An online survey methodology was employed to collect data including personality traits (NEO-FFI-R), impulsivity (BIS-11), sensation seeking (ImpSS), and demographic information. The data set contained information on the consumption of 18 central nervous system psychoactive drugs. Correlation analysis using a relative information gain model demonstrates the existence of a group of drugs (amphetamines, cannabis, cocaine, ecstasy, legal highs, LSD, and magic mushrooms) with strongly correlated consumption. An exhaustive search was performed to select the most effective subset of input features and data mining methods to classify users and non-users for each drug. A number of classification methods were employed (decision tree, random forest, k-nearest neighbours, linear discriminant analysis, Gaussian mixture, probability density function estimation, logistic regression, and naive Bayes) and the most effective method selected for each drug. The quality of classification was surprisingly high. The best results with sensitivity and specificity being greater than 75% were achieved for cannabis, crack, ecstasy, legal highs, LSD, and volatile substance abuse. Sensitivity and specificity greater than 70% were achieved for amphetamines, amyl nitrite, benzodiazepines, chocolate, caffeine, heroin, ketamine, methadone, and nicotine. The poorest result was obtained for prediction of alcohol consumption.

Clustered intergenic region sequences as predictors of factor H Binding Protein expression patterns and for assessing Neisseria meningitidis strain coverage by meningococcal vaccines

Factor H binding protein (fHbp) is a major protective antigen in 4C-MenB (Bexsero®) and Trumenba®, two serogroup B meningococcal vaccines, wherein expression level is a determinant of protection. Examination of promoter-containing intergenic region (IGR) sequences indicated that nine fHbp IGR alleles covered 92% of 1,032 invasive meningococcal strains with variant 1 fHbp alleles. Relative expression values for fHbp were determined for 79 meningococcal isolates covering ten IGR alleles by quantitative reverse transcriptase polymerase chain reaction (qRT PCR). Derivation of expression clusters of IGR sequences by linear regression identified five expression clusters with five nucleotides and one insertion showing statistically associations with differences in expression level. Sequence analysis of 273 isolates examined by the Meningococcal Antigen Typing Scheme, a sandwich ELISA, found that coverage depended on the IGR expression cluster and vaccine peptide homology combination. Specific fHbp peptide-IGR expression cluster combinations were designated as ‘at risk’ for coverage by 4C-MenB and were detected in multiple invasive meningococcal disease cases confirmed by PCR alone and occurring in partially-vaccinated infants. We conclude that sequence-based analysis of IGR sequences is informative for assessing protein expression and has utility for culture-independent assessments of strain coverage by protein-based vaccines.

Theoretical aspects of peptide imprinting: screening of MIP (virtual) binding sites for their interactions with amino acids, di- and tripeptides

ABSTRACTMolecular modelling and computational approaches were used to design (virtual) molecularly imprinted binding sited for 170 amino acids, dipeptides and tripeptides. Analysis of the binding energy of ligands to their corresponding virtual binding sites revealed a direct correlation between size of the ligand and its binding affinity. Only tripeptides were capable of forming binding sites in molecularly imprinted polymers (MIPs) that are capable, in theory, of binding the corresponding targets at micromolar concentrations. No appreciable specificity was demonstrated in binding of virtual binding sites and corresponding templates. It is possible to conclude that although tripeptide sequences are sufficiently long to form MIPs with relatively high affinity, the sequence of peptide epitopes should be substantially longer that three amino acid residues to ensure specificity of imprinted sites. This consideration will be useful for the design of highly efficient MIPs for proteins.