As noncommunicable diseases (NCDs) pose a significant global health burden, identifying effective diagnostic and predictive markers for these diseases is of paramount importance. Epigenetic modifications, such as DNA methylation, have emerged as potential indicators for NCDs. These have previously been exploited in other contexts within the framework of neural network models that capture complex relationships within the data. Applications of neural networks have led to significant breakthroughs in various biological or biomedical fields but these have not yet been effectively applied to NCD modeling. This is, in part, due to limited datasets that are not amenable to building of robust neural network models. In this work, we leveraged a neural network trained on one class of NCDs, cancer, as the basis for a transfer learning approach to non-cancer NCD modeling. Our results demonstrate promising performance of the model in predicting three NCDs, namely, arthritis, asthma, and schizophrenia, for the respective blood samples, with an overall accuracy (f-measure) of 94.5%. Furthermore, a concept based explanation method called Testing with Concept Activation Vectors (TCAV) was used to investigate the importance of the sample sources and understand how future training datasets for multiple NCD models may be improved. Our findings highlight the effectiveness of transfer learning in developing accurate diagnostic and predictive models for NCDs.

BACKGROUND: Despite remarkable advances in cancer research, cancer remains one of the leading causes of death worldwide. Early detection of cancer and localization of the tissue of its origin are key to effective treatment. Here, we leverage technological advances in machine learning or artificial intelligence to design a novel framework for cancer diagnostics. Our proposed framework detects cancers and their tissues of origin using a unified model of cancers encompassing 33 cancers represented in The Cancer Genome Atlas (TCGA). Our model exploits the learned features of different cancers reflected in the respective dysregulated epigenomes, which arise early in carcinogenesis and differ remarkably between different cancer types or subtypes, thus holding a great promise in early cancer detection. RESULTS: Our comprehensive assessment of the proposed model on the 33 different tissues of origin demonstrates its ability to detect and classify cancers to a high accuracy (> 99% overall F-measure). Furthermore, our model distinguishes cancers from pre-cancerous lesions to metastatic tumors and discriminates between hypomethylation changes due to age related epigenetic drift and true cancer. CONCLUSIONS: Beyond detection of primary cancers, our proposed computational model also robustly detects tissues of origin of secondary cancers, including metastatic cancers, second primary cancers, and cancers of unknown primaries. Our assessment revealed the ability of this model to characterize pre-cancer samples, a significant step forward in early cancer detection. Deployed broadly this model can deliver accurate diagnosis for a greatly expanded target patient population.

BACKGROUND: Genomic Islands (GIs) are clusters of genes that are mobilized through horizontal gene transfer. GIs play a pivotal role in bacterial evolution as a mechanism of diversification and adaptation to different niches. Therefore, identification and characterization of GIs in bacterial genomes is important for understanding bacterial evolution. However, quantifying GIs is inherently difficult, and the existing methods suffer from low prediction accuracy and precision-recall trade-off. Moreover, several of them are supervised in nature, and thus, their applications to newly sequenced genomes are riddled with their dependency on the functional annotation of existing genomes. RESULTS: We present SSG-LUGIA, a completely automated and unsupervised approach for identifying GIs and horizontally transferred genes. SSG-LUGIA is a novel method based on unsupervised anomaly detection technique, accompanied by further refinement using cues from signal processing literature. SSG-LUGIA leverages the atypical compositional biases of the alien genes to localize GIs in prokaryotic genomes. SSG-LUGIA was assessed on a large benchmark dataset `IslandPick' and on a set of 15 well-studied genomes in the literature and followed by a thorough analysis on the well-understood Salmonella typhi CT18 genome. Furthermore, the efficacy of SSG-LUGIA in identifying horizontally transferred genes was evaluated on two additional bacterial genomes, namely, those of Corynebacterium diphtheria NCTC13129 and Pseudomonas aeruginosa LESB58. SSG-LUGIA was examined on draft genomes and was demonstrated to be efficient as an ensemble method. CONCLUSIONS: Our results indicate that SSG-LUGIA achieved superior performance in comparison to frequently used existing methods. Importantly, it yielded a better trade-off between precision and recall than the existing methods. Its nondependency on the functional annotation of genomes makes it suitable for analyzing newly sequenced, yet uncharacterized genomes. Thus, our study is a significant advance in identification of GIs and horizontally transferred genes. SSG-LUGIA is available as an open source software at https://nibtehaz.github.io/SSG-LUGIA/.

Recently, the frequency of observing bacterial strains without known genetic components underlying phenotypic resistance to antibiotics has increased. There are several strains of bacteria lacking known resistance genes; however, they demonstrate resistance phenotype to drugs of that family. Although such strains are fewer compared to the overall population, they pose grave emerging threats to an already heavily challenged area of antimicrobial resistance (AMR), where death tolls have reached ~700 000 per year and a grim projection of ~10 million deaths per year by 2050 looms. Considering the fact that development of novel antibiotics is not keeping pace with the emergence and dissemination of resistance, there is a pressing need to decipher yet unknown genetic mechanisms of resistance, which will enable developing strategies for the best use of available interventions and show the way for the development of new drugs. In this study, we present a machine learning framework to predict novel AMR factors that are potentially responsible for resistance to specific antimicrobial drugs. The machine learning framework utilizes whole-genome sequencing AMR genetic data and antimicrobial susceptibility testing phenotypic data to predict resistance phenotypes and rank AMR genes by their importance in discriminating the resistance from the susceptible phenotypes. In summary, we present here a bioinformatics framework for training machine learning models, evaluating their performances, selecting the best performing model(s) and finally predicting the most important AMR loci for the resistance involved.

MOTIVATION: Alignment-free, stochastic models derived from k-mer distributions representing reference genome sequences have a rich history in the classification of DNA sequences. In particular, the variants of Markov models have previously been used extensively. Higher-order Markov models have been used with caution, perhaps sparingly, primarily because of the lack of enough training data and computational power. Advances in sequencing technology and computation have enabled exploitation of the predictive power of higher-order models. We, therefore, revisited higher-order Markov models and assessed their performance in classifying metagenomic sequences. RESULTS: Comparative assessment of higher-order models (HOMs, 9th order or higher) with interpolated Markov model, interpolated context model and lower-order models (8th order or lower) was performed on metagenomic datasets constructed using sequenced prokaryotic genomes. Our results show that HOMs outperform other models in classifying metagenomic fragments as short as 100 nt at all taxonomic ranks, and at lower ranks when the fragment size was increased to 250 nt. HOMs were also found to be significantly more accurate than local alignment which is widely relied upon for taxonomic classification of metagenomic sequences. A novel software implementation written in C++ performs classification faster than the existing Markovian metagenomic classifiers and can therefore be used as a standalone classifier or in conjunction with existing taxonomic classifiers for more robust classification of metagenomic sequences. AVAILABILITY AND IMPLEMENTATION: The software has been made available at https://github.com/djburks/SMM. CONTACT: [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Precision medicine is rapidly emerging as a strategy to tailor medical treatment to a small group or even individual patients based on their genetics, environment and lifestyle. Precision medicine relies heavily on developments in systems biology and omics disciplines, including metabolomics. Combination of metabolomics with sophisticated bioinformatics analysis and mathematical modeling has an extreme power to provide a metabolic snapshot of the patient over the course of disease and treatment or classifying patients into subpopulations and subgroups requiring individual medical treatment. Although a powerful approach, metabolomics have certain limitations in technology and bioinformatics. We will review various aspects of metabolomics technology and bioinformatics, from data generation, bioinformatics analysis, data fusion and mathematical modeling to data management, in the context of precision medicine.

Parametric methods for identifying laterally transferred genes exploit the directional mutational biases unique to each genome. Yet the development of new, more robust methods--as well as the evaluation and proper implementation of existing methods--relies on an arbitrary assessment of performance using real genomes, where the evolutionary histories of genes are not known. We have used the framework of a generalized hidden Markov model to create artificial genomes modeled after genuine genomes. To model a genome, "core" genes--those displaying patterns of mutational biases shared among large numbers of genes--are identified by a novel gene clustering approach based on the Akaike information criterion. Gene models derived from multiple "core" gene clusters are used to generate an artificial genome that models the properties of a genuine genome. Chimeric artificial genomes--representing those having experienced lateral gene transfer--were created by combining genes from multiple artificial genomes, and the performance of the parametric methods for identifying "atypical" genes was assessed directly. We found that a hidden Markov model that included multiple gene models, each trained on sets of genes representing the range of genotypic variability within a genome, could produce artificial genomes that mimicked the properties of genuine genomes. Moreover, different methods for detecting foreign genes performed differently--i.e., they had different sets of strengths and weaknesses--when identifying atypical genes within chimeric artificial genomes.

In this paper, we review developments in probabilistic methods of gene recognition in prokaryotic genomes with the emphasis on connections to the general theory of hidden Markov models (HMM). We show that the Bayesian method implemented in GeneMark, a frequently used gene-finding tool, can be augmented and reintroduced as a rigorous forward-backward (FB) algorithm for local posterior decoding described in the HMM theory. Another earlier developed method, prokaryotic GeneMark.hmm, uses a modification of the Viterbi algorithm for HMM with duration to identify the most likely global path through hidden functional states given the DNA sequence. GeneMark and GeneMark.hmm programs are worth using in concert for analysing prokaryotic DNA sequences that arguably do not follow any exact mathematical model. The new extension of GeneMark using the FB algorithm was implemented in the software program GeneMark.fba. Given the DNA sequence, this program determines an a posteriori probability for each nucleotide to belong to coding or non-coding region. Also, for any open reading frame (ORF), it assigns a score defined as a probabilistic measure of all paths through hidden states that traverse the ORF as a coding region. The prediction accuracy of GeneMark.fba determined in our tests was compared favourably to the accuracy of the initial (standard) GeneMark program. Comparison to the prokaryotic GeneMark.hmm has also demonstrated a certain, yet species-specific, degree of improvement in raw gene detection, ie detection of correct reading frame (and stop codon). The accuracy of exact gene prediction, which is concerned about precise prediction of gene start (which in a prokaryotic genome unambiguously defines the reading frame and stop codon, thus, the whole protein product), still remains more accurate in GeneMarkS, which uses more elaborate HMM to specifically address this task.

MOTIVATION: Markov chain models of DNA sequences have frequently been used in gene finding algorithms. Performance of the algorithm critically depends on the model structure and parameters. Still, the issue of choosing the model structure has not been studied with sufficient attention. RESULTS: We have assessed performance of several types of Markov chain models, both fixed order (FO) models and models with interpolation, within the framework of the GeneMark algorithm. The performance was measured in two ways: (i) the accuracy of detection of protein-coding potential in artificial DNA sequences and (ii) the accuracy of identifying genes in real prokaryotic genomes. We observed that the models built by deleted interpolation (DI) slightly outperformed other models in detecting protein-coding potential in artificial DNA sequences with GC content in the medium range and also in detecting genes in real genomes with medium GC content. For artificial and real genomic DNA with high or low GC content, we observed that the models built by DI were in some cases slightly outperformed by FO models.