BACKGROUND: In unforeseen situations, such as nuclear power plant's or civilian radiation accidents, there is a need for effective and computationally inexpensive methods to determine the expression level of a selected gene panel, allowing for rough dose estimates in thousands of donors. The new generation in-situ mapper, fast and of low energy consumption, working at the level of single nanopore output, is in demand. We aim to create a sequence identification tool that utilizes natural language processing techniques and ensures a high level of negative predictive value (NPV) compared to the classical approach. RESULTS: The training dataset consisted of RNA sequencing data from 6 samples. Multiple natural language processing models were examined, differing in the type of dictionary components (word length, step, context) as well as the encoding length and number of sequences required for algorithm training. The best configuration analyses the entire sequence and uses a word length of 3 base pairs with one-word neighbor on each side. For the considered FDXR gene, the achieved mean balanced accuracy (BACC) was 98.29% and NPV was 99.25%, compared to minimap2's performance in a cross-validation scenario. The next stage focused on exploring the dictionary components and attempting to optimize it, employing statistical techniques as well as those relying on the explainability of the decisions made. Reducing the dictionary from 1024 to 145 changed BACC to 96.49% and the NPV to 98.15%. Obtained model, validated on an external independent genome sequencing dataset, gave NPV of 99.64% for complete and 95.87% for reduced dictionary. The salmon-estimated read counts differed from the classical approach on average by 3.48% for the complete dictionary and by 5.82% for the reduced one. CONCLUSIONS: We conclude that for long Oxford nanopore reads, a natural language processing-based approach can reliably replace classical mapping when there is a need for fast, reliable and energy and computationally efficient targeted mapping of a pre-defined subset of transcripts. The developed model can be easily retrained to identify selected transcripts and/or work with various long-read sequencing techniques. Our results of the study clearly demonstrate the potential of applying techniques known from classical text processing to nucleotide sequences.

BACKGROUND: Investigating molecular heterogeneity provides insights into tumour origin and metabolomics. The increasing amount of data gathered makes manual analyses infeasible-therefore, automated unsupervised learning approaches are utilised for discovering tissue heterogeneity. However, automated analyses require experience setting the algorithms' hyperparameters and expert knowledge about the analysed biological processes. Moreover, feature engineering is needed to obtain valuable results because of the numerous features measured. RESULTS: We propose DiviK: a scalable stepwise algorithm with local data-driven feature space adaptation for segmenting high-dimensional datasets. The algorithm is compared to the optional solutions (regular k-means, spatial and spectral approaches) combined with different feature engineering techniques (None, PCA, EXIMS, UMAP, Neural Ions). Three quality indices: Dice Index, Rand Index and EXIMS score, focusing on the overall composition of the clustering, coverage of the tumour region and spatial cluster consistency, are used to assess the quality of unsupervised analyses. Algorithms were validated on mass spectrometry imaging (MSI) datasets-2D human cancer tissue samples and 3D mouse kidney images. DiviK algorithm performed the best among the four clustering algorithms compared (overall quality score 1.24, 0.58 and 162 for d(0, 0, 0), d(1, 1, 1) and the sum of ranks, respectively), with spectral clustering being mostly second. Feature engineering techniques impact the overall clustering results less than the algorithms themselves (partial [Formula: see text] effect size: 0.141 versus 0.345, Kendall's concordance index: 0.424 versus 0.138 for d(0, 0, 0)). CONCLUSIONS: DiviK could be the default choice in the exploration of MSI data. Thanks to its unique, GMM-based local optimisation of the feature space and deglomerative schema, DiviK results do not strongly depend on the feature engineering technique applied and can reveal the hidden structure in a tissue sample. Additionally, DiviK shows high scalability, and it can process at once the big omics data with more than 1.5 mln instances and a few thousand features. Finally, due to its simplicity, DiviK is easily generalisable to an even more flexible framework. Therefore, it is helpful for other -omics data (as single cell spatial transcriptomic) or tabular data in general (including medical images after appropriate embedding). A generic implementation is freely available under Apache 2.0 license at https://github.com/gmrukwa/divik .

Mass cytometry as an advanced single-cell analysis technology can produce high-dimensional data consisting of millions of cells and more than 50 features. Therefore the cell subtypes identification is difficult and impossible to be done manually. Each step of the analysis affect the results and may cause a loss of rare sub-populations of interest. One of the first steps in the analysis is pre-gating which involves filtering out unwanted measurements like debris or doublets. The existing semi-automated solutions for pre-gating require some parameters to be set which may lead to different results. Moreover, the tools often use downsampling from millions to thousands of cells. Despite the existing methods, there is still a need for a fully automated tool that will be independent of sample size. In the study, we developed a solution based on Gaussian Mixture Model (GMM) decomposition and grouping of its components into clusters. Based on the clusters we propose filtration criteria that identify measurements to be removed from the analysis. The algorithm was validated on two independent public datasets. The results are promising and reproducible, leaving intact, live cells that can be further analyzed.

Data filtering based on removing non-informative features, with unchanged signal between compared experimental conditions, can significantly increase sensitivity of methods used to detect differentially expressed genes or other molecular components measured in high-throughput biological experiments. Criteria for data filtering can be stated on the basis of averages or variances of signal levels across samples. The crucial parts of feature filtering are selection of filter type and cut-off threshold, which are specific to the particular dataset. In this paper, we present an algorithm and a stand-alone application, GaMRed, for adaptive filtering insignificant features in high-throughput data, based on Gaussian mixture decomposition. We have tested the performance of our algorithm using datasets from three different high-throughput biological experiments. We estimated the number of differentially expressed features after applying multiple testing correction and performed functional analysis of obtained features using Gene Ontology terms. Also, we checked if the control of false discovery rate and family-wise error rate after applying feature filtering remains at appropriate level. GaMRed is fast, automatic, and does not require expert knowledge in parameter tuning. The algorithm increases sensitivity of methods used to find differentially expressed features and biological validity of the findings. The program can be downloaded from: http://zaed.aei.polsl.pl/index.php/pl/oprogramowanie-zaed.

Breast cancer, a frequent cause of mortality among women, is a heterogeneous disease with multiple subtypes, which identification has a great impact on the prognosis and the choice of therapy. Various attempts are being made worldwide to classify breast tumors and to discover differences between subtypes. The purpose of this study is to divide a diverse group of patients with a multiple k-means clustering approach in order to investigate the complex structure of the dataset. Primary tumor samples collected from 859 females within The Cancer Genome Atlas Breast Invasive Carcinoma project served as the material for an RNA-Seq experiment. Expression levels measured with the RNA-Seq procedure were downloaded from the Genomic Data Commons Data Portal and used for this study. The proposed multilevel clustering algorithm is based on the idea of treating each resulting cluster as an input for a new k-means clustering as long as a given cluster is not too small or integrated. Each clustering is preceded with feature filtration and selection, which are based on the Gaussian Mixture Model decomposition of gene expression level amplitude or variance, respectively. The gap statistic is used for the selection of the optimal number of clusters as well as for deciding whether a group should be divided. Results were compared to the subtype labels either obtained from the PAM50 classifier or based on the HER2 and hormone-receptor markers. The performed multilevel clustering produced a large number of small clusters. During the first division five clusters were obtained, three of them containing almost only hormone-receptor positive (HR+) tumors. One cluster consisted of similar numbers of HER2-enriched, triple-negative breast cancer (TNBC), and HR+ cases. Obtained results confirm a great heterogeneity of breast cancer. However, they also suggest that commonly used breast tumor classification corresponding to the receptor markers may not entirely reflect the complexity of subtypes of this disease. The proposed algorithm gives a comprehensive insight into the hierarchical composition of dataset and allows for identification of genes influencing each division, which may be used for further investigation of the heterogeneity of breast cancer.

MOTIVATION: In contemporary biological experiments, bias, which interferes with the measurements, requires attentive processing. Important sources of bias in high-throughput biological experiments are batch effects and diverse methods towards removal of batch effects have been established. These include various normalization techniques, yet many require knowledge on the number of batches and assignment of samples to batches. Only few can deal with the problem of identification of batch effect of unknown structure. For this reason, an original batch identification algorithm through dynamical programming is introduced for omics data that may be sorted on a timescale. RESULTS: BatchI algorithm is based on partitioning a series of high-throughput experiment samples into sub-series corresponding to estimated batches. The dynamic programming method is used for splitting data with maximal dispersion between batches, while maintaining minimal within batch dispersion. The procedure has been tested on a number of available datasets with and without prior information about batch partitioning. Datasets with a priori identified batches have been split accordingly, measured with weighted average Dice Index. Batch effect correction is justified by higher intra-group correlation. In the blank datasets, identified batch divisions lead to improvement of parameters and quality of biological information, shown by literature study and Information Content. The outcome of the algorithm serves as a starting point for correction methods. It has been demonstrated that omitting the essential step of batch effect control may lead to waste of valuable potential discoveries. AVAILABILITY AND IMPLEMENTATION: The implementation is available within the BatchI R package at http://zaed.aei.polsl.pl/index.php/pl/111-software. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Analysis of gene set (GS) enrichment is an essential part of functional omics studies. Here, we complement the established evaluation metrics of GS enrichment algorithms with a novel approach to assess the practical reproducibility of scientific results obtained from GS enrichment tests when applied to related data from different studies. RESULTS: We evaluated eight established and one novel algorithm for reproducibility, sensitivity, prioritization, false positive rate and computational time. In addition to eight established algorithms, we also included Coincident Extreme Ranks in Numerical Observations (CERNO), a flexible and fast algorithm based on modified Fisher P-value integration. Using real-world datasets, we demonstrate that CERNO is robust to ranking metrics, as well as sample and GS size. CERNO had the highest reproducibility while remaining sensitive, specific and fast. In the overall ranking Pathway Analysis with Down-weighting of Overlapping Genes, CERNO and over-representation analysis performed best, while CERNO and GeneSetTest scored high in terms of reproducibility. AVAILABILITY AND IMPLEMENTATION: tmod package implementing the CERNO algorithm is available from CRAN (cran.r-project.org/web/packages/tmod/index.html) and an online implementation can be found at http://tmod.online/. The datasets analyzed in this study are widely available in the KEGGdzPathwaysGEO, KEGGandMetacoreDzPathwaysGEO R package and GEO repository. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: There exist many methods for describing the complex relation between changes of gene expression in molecular pathways or gene ontologies under different experimental conditions. Among them, Gene Set Enrichment Analysis seems to be one of the most commonly used (over 10,000 citations). An important parameter, which could affect the final result, is the choice of a metric for the ranking of genes. Applying a default ranking metric may lead to poor results. METHODS AND RESULTS: In this work 28 benchmark data sets were used to evaluate the sensitivity and false positive rate of gene set analysis for 16 different ranking metrics including new proposals. Furthermore, the robustness of the chosen methods to sample size was tested. Using k-means clustering algorithm a group of four metrics with the highest performance in terms of overall sensitivity, overall false positive rate and computational load was established i.e. absolute value of Moderated Welch Test statistic, Minimum Significant Difference, absolute value of Signal-To-Noise ratio and Baumgartner-Weiss-Schindler test statistic. In case of false positive rate estimation, all selected ranking metrics were robust with respect to sample size. In case of sensitivity, the absolute value of Moderated Welch Test statistic and absolute value of Signal-To-Noise ratio gave stable results, while Baumgartner-Weiss-Schindler and Minimum Significant Difference showed better results for larger sample size. Finally, the Gene Set Enrichment Analysis method with all tested ranking metrics was parallelised and implemented in MATLAB, and is available at https://github.com/ZAEDPolSl/MrGSEA . CONCLUSIONS: Choosing a ranking metric in Gene Set Enrichment Analysis has critical impact on results of pathway enrichment analysis. The absolute value of Moderated Welch Test has the best overall sensitivity and Minimum Significant Difference has the best overall specificity of gene set analysis. When the number of non-normally distributed genes is high, using Baumgartner-Weiss-Schindler test statistic gives better outcomes. Also, it finds more enriched pathways than other tested metrics, which may induce new biological discoveries.

Although there are several data analysis frameworks, both commercial and open source, supporting the detection of tumours on nuclear magnetic resonance (NMR) sequences, none of them gives satisfactory results in the case of low volume tumors. The majority of the frameworks require the detailed analysis of at least two sequences of the examined sample, or give sample specific thresholds distinguishing between the tumor and subtypes of healthy tissue. In this paper, we present a novel algorithm for the automated estimation of tumor specific cut-off values in the domain of the apparent diffusion coefficient (ADC). Once the cut-off characteristics for a particular type of tumor is estimated, their further usage on other independent samples does not require any calculations except for an easy thresholding. The proposed methodology is a combination of classical decomposition of ADC distribution into a Gaussian mixture model (GMM) with k-means clustering subsequently performed on the parameters of mixture model components, leading to the identification of ADC distributions for every tissue type. The maximum conditional probability criterion gives the final threshold estimate. The developed signal analysis pipeline was applied to the problem of Glioblastoma Multiforme grade IV brain tumor segmentation, with a dataset of 119 randomly chosen ADC maps and Leave-One-Out cross-validation procedure for population error estimate. Additionally, a comparison to standard GMM based tumor segmentation algorithms as well as to three other automated segmentation methods was performed and the obtained tumor regions were referenced to the segmentation done by a human expert. The results demonstrate the average MiMSeg similarity to the expert-curated decision measured by the Dice coefficient as equal to 89.2% (with 95% confidence interval 87.7 ÷ 90.6). The MiMSeg algorithm significantly outperforms other techniques in the case of small tumors (of volume less than 10%), obtaining similarity to the expert-curated decision at the level 86.7%, with 44.9% obtained by standard GMM, 79.0% by Self-Organising-Maps algorithm, 68.7% by Murakami’s algorithm, and 78.2% by Kang’s method.

We propose a new method that employs transfer learning techniques to effectively correct sampling selection errors introduced by sparse annotations during supervised learning for automated tumor segmentation. The practicality of current learning-based automated tissue classification approaches is severely impeded by their dependency on manually segmented training databases that need to be recreated for each scenario of application, site, or acquisition setup. The comprehensive annotation of reference datasets can be highly labor-intensive, complex, and error-prone. The proposed method derives high-quality classifiers for the different tissue classes from sparse and unambiguous annotations and employs domain adaptation techniques for effectively correcting sampling selection errors introduced by the sparse sampling. The new approach is validated on labeled, multi-modal MR images of 19 patients with malignant gliomas and by comparative analysis on the BraTS 2013 challenge data sets. Compared to training on fully labeled data, we reduced the time for labeling and training by a factor greater than 70 and 180 respectively without sacrificing accuracy. This dramatically eases the establishment and constant extension of large annotated databases in various scenarios and imaging setups and thus represents an important step towards practical applicability of learning-based approaches in tissue classification.

BACKGROUND: DNA microarrays are used for discovery of genes expressed differentially between various biological conditions. In microarray experiments the number of analyzed samples is often much lower than the number of genes (probe sets) which leads to many false discoveries. Multiple testing correction methods control the number of false discoveries but decrease the sensitivity of discovering differentially expressed genes. Concerning this problem, filtering methods for improving the power of detection of differentially expressed genes were proposed in earlier papers. These techniques are two-step procedures, where in the first step some pool of non-informative genes is removed and in the second step only the pool of the retained genes is used for searching for differentially expressed genes. RESULTS: A very important parameter to choose is the proportion between the sizes of the pools of removed and retained genes. A new method, which we propose, allow to determine close to optimal threshold values for sample means and sample variances for gene filtering. The method is adaptive and based on the decomposition of the histogram of gene expression means or variances into mixture of Gaussian components. CONCLUSIONS: By performing analyses of several publicly available datasets and simulated datasets we demonstrate that our adaptive method increases sensitivity of finding differentially expressed genes compared to previous methods of filtering microarray data based on using fixed threshold values.

Abstract The article presents a method of protein matrix‐assisted laser desorption and ionization‐time of flight (TOF) spectra analysis. The method performs peaks detection. Spectra are analysed with Gaussian mixture decomposition. The results obtained are used for peaks identification purposes. The concept of the method is that a single peak is represented by one Gaussian distribution. The expectation‐maximization algorithm and maximum likelihood rule are used for spectra processing. The analysis can be done for a set of spectra with use of the mean spectrum, or it may be performed for a single spectrum at a time. The number of mixture‐model components is estimated by the Bayesian information criterion. Before the main analysis, a few pre‐processing steps need to be done. Spectra should be subjected to calibration, normalization, denoising, baseline correction, etc. The aim of the work is to identify peptides in the analysed sample on the basis of the parameters of the mixture model and to find differences between spectra in the analysed set.