A systematic review of the application of machine learning in the detection and classification of transposable elements

Teaching transposon classification as a means to crowd source the curation of repeat annotation - a tardigrade perspective

BACKGROUND

The advancement of sequencing technologies results in the rapid release of hundreds of new genome assemblies a year providing unprecedented resources for the study of genome evolution. Within this context, the significance of in-depth analyses of repetitive elements, transposable elements (TEs) in particular, is increasingly recognized in understanding genome evolution. Despite the plethora of available bioinformatic tools for identifying and annotating TEs, the phylogenetic distance of the target species from a curated and classified database of repetitive element sequences constrains any automated annotation effort. Moreover, manual curation of raw repeat libraries is deemed essential due to the frequent incompleteness of automatically generated consensus sequences.

RESULTS

Here, we present an example of a crowd-sourcing effort aimed at curating and annotating TE libraries of two non-model species built around a collaborative, peer-reviewed teaching process. Manual curation and classification are time-consuming processes that offer limited short-term academic rewards and are typically confined to a few research groups where methods are taught through hands-on experience. Crowd-sourcing efforts could therefore offer a significant opportunity to bridge the gap between learning the methods of curation effectively and empowering the scientific community with high-quality, reusable repeat libraries.

CONCLUSIONS

The collaborative manual curation of TEs from two tardigrade species, for which there were no TE libraries available, resulted in the successful characterization of hundreds of new and diverse TEs in a reasonable time frame. Our crowd-sourcing setting can be used as a teaching reference guide for similar projects: A hidden treasure awaits discovery within non-model organisms.

Machine learning applications on intratumoral heterogeneity in glioblastoma using single-cell RNA sequencing data

Artificial intelligence is revolutionizing all fields that affect people's lives and health. One of the most critical applications is in the study of tumors. It is the case of glioblastoma (GBM) that has behaviors that need to be understood to develop effective therapies. Due to advances in single-cell RNA sequencing (scRNA-seq), it is possible to understand the cellular and molecular heterogeneity in the GBM. Given that there are different cell groups in these tumors, there is a need to apply Machine Learning (ML) algorithms. It will allow extracting information to understand how cancer changes and broaden the search for effective treatments. We proposed multiple comparisons of ML algorithms to classify cell groups based on the GBM scRNA-seq data. This broad comparison spectrum can show the scientific-medical community which models can achieve the best performance in this task. In this work are classified the following cell groups: Tumor Core (TC), Tumor Periphery (TP) and Normal Periphery (NP), in binary and multi-class scenarios. This work presents the biomarker candidates found for the models with the best results. The analyses presented here allow us to verify the biomarker candidates to understand the genetic characteristics of GBM, which may be affected by a suitable identification of GBM heterogeneity. This work obtained for the four scenarios covered cross-validation results of $93.03\% \pm 5.37\%$, $97.42\% \pm 3.94\%$, $98.27\% \pm 1.81\%$ and $93.04\% \pm 6.88\%$ for the classification of TP versus TC, TP versus NP, NP versus TP and TC (TPC) and NP versus TP versus TC, respectively.

The distribution and characteristic of two transposable elements in the genome of Cydia pomonella granulovirus and codling moth

Baculoviruses are capable to acquire insect host transposable elements (TEs) in their genomes and are hypothesized as possible vectors of insect transposons between Lepidopteran species. Here, we investigated the host origin of two TEs, namely the Tc1/mariner-like element TCp3.2 and a 0.7 kbp insertion sequence (IS07), found in the genome of different isolates of Cydia pomonella granulovirus (CpGV), a member of the Betabaculovirus genus. The sequences of both TEs were searched for in the full genome sequence database of codling moth (CM, Cydia pomonella L.). A total of eleven TCp3.2 TE copies and 76 copies of the IS07 fragments were identified in the CM genome. These TEs were distributed over the 22 autosomes and the Z chromosome (chr1) of CM, except chr6, chr12, chr16, chr23, chr27 and the W chromosome (chr29). TCp3.2 copies with two transposase genes in opposite direction, representing a novel feature, were identified on chr10 and chr18. The TCp3.2 transposase was characterized by DD41D motif of classic Tc1/mariner transposons, consisting of DNA-binding domain, catalytic domain and nuclear localization signal (NLS). Transcription analyses of uninfected and CpGV-infected CM larvae suggested a doubling of the TCp3.2 transposase transcription rate in virus infected larvae. Furthermore, IS07 insertion into the CpGV genome apparently added new transcription initiation sites to the viral genome. The global analysis of the distribution of two TEs in the genome of CM addressed the influx of mobile TEs from CM to CpGV, a genetic process that contributes to the population diversity of baculoviruses.

Inpactor2: a software based on deep learning to identify and classify LTR-retrotransposons in plant genomes

LTR-retrotransposons are the most abundant repeat sequences in plant genomes and play an important role in evolution and biodiversity. Their characterization is of great importance to understand their dynamics. However, the identification and classification of these elements remains a challenge today. Moreover, current software can be relatively slow (from hours to days), sometimes involve a lot of manual work and do not reach satisfactory levels in terms of precision and sensitivity. Here we present Inpactor2, an accurate and fast application that creates LTR-retrotransposon reference libraries in a very short time. Inpactor2 takes an assembled genome as input and follows a hybrid approach (deep learning and structure-based) to detect elements, filter partial sequences and finally classify intact sequences into superfamilies and, as very few tools do, into lineages. This tool takes advantage of multi-core and GPU architectures to decrease execution times. Using the rice genome, Inpactor2 showed a run time of 5 minutes (faster than other tools) and has the best accuracy and F1-Score of the tools tested here, also having the second best accuracy and specificity only surpassed by EDTA, but achieving 28% higher sensitivity. For large genomes, Inpactor2 is up to seven times faster than other available bioinformatics tools.

Automatic curation of LTR retrotransposon libraries from plant genomes through machine learning

Transposable elements are mobile sequences that can move and insert themselves into chromosomes, activating under internal or external stimuli, giving the organism the ability to adapt to the environment. Annotating transposable elements in genomic data is currently considered a crucial task to understand key aspects of organisms such as phenotype variability, species evolution, and genome size, among others. Because of the way they replicate, LTR retrotransposons are the most common transposable elements in plants, accounting in some cases for up to 80% of all DNA information. To annotate these elements, a reference library is usually created, a curation process is performed, eliminating TE fragments and false positives and then annotated in the genome using the homology method. However, the curation process can take weeks, requires extensive manual work and the execution of multiple time-consuming bioinformatics software. Here, we propose a machine learning-based approach to perform this process automatically on plant genomes, obtaining up to 91.18% F1-score. This approach was tested with four plant species, obtaining up to 93.6% F1-score (Oryza granulata) in only 22.61 s, where bioinformatics methods took approximately 6 h. This acceleration demonstrates that the ML-based approach is efficient and could be used in massive sequencing projects.

Finding and Characterizing Repeats in Plant Genomes

Plant genomes contain a particularly high proportion of repeated structures of various types. This chapter proposes a guided tour of the available software that can help biologists to scan automatically for these repeats in sequence data or check hypothetical models intended to characterize their structures. Since transposable elements (TEs) are a major source of repeats in plants, many methods have been used or developed for this broad class of sequences. They are representative of the range of tools available for other classes of repeats and we have provided two sections on this topic (for the analysis of genomes or directly of sequenced reads), as well as a selection of the main existing software. It may be hard to keep up with the profusion of proposals in this dynamic field and the rest of the chapter is devoted to the foundations of an efficient search for repeats and more complex patterns. We first introduce the key concepts of the art of indexing and mapping or querying sequences. We end the chapter with the more prospective issue of building models of repeat families. We present the Machine Learning approach first, seeking to build predictors automatically for some families of ET, from a set of sequences known to belong to this family. A second approach, the linguistic (or syntactic) approach, allows biologists to describe themselves and check the validity of models of their favorite repeat family.

Machine Learning Algorithms to Detect Sex in Myocardial Perfusion Imaging

Myocardial perfusion imaging (MPI) is an essential tool used to diagnose and manage patients with suspected or known coronary artery disease. Additionally, the General Data Protection Regulation (GDPR) represents a milestone about individuals' data security concerns. On the other hand, Machine Learning (ML) has had several applications in the most diverse knowledge areas. It is conceived as a technology with huge potential to revolutionize health care. In this context, we developed ML models to evaluate their ability to distinguish an individual's sex from MPI assessment. We used 260 polar maps (140 men/120 women) to train ML algorithms from a database of patients referred to a university hospital for clinically indicated MPI from January 2016 to December 2018. We tested 07 different ML models, namely, Classification and Regression Tree (CART), Naive Bayes (NB), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Adaptive Boosting (AB), Random Forests (RF) and, Gradient Boosting (GB). We used a cross-validation strategy. Our work demonstrated that ML algorithms could perform well in assessing the sex of patients undergoing myocardial scintigraphy exams. All the models had accuracy greater than 82%. However, only SVM achieved 90%. KNN, RF, AB, GB had, respectively, 88, 86, 85, 83%. Accuracy standard deviation was lower in KNN, AB, and RF (0.06). SVM and RF had had the best area under the receiver operating characteristic curve (0.93), followed by GB (0.92), KNN (0.91), AB, and NB (0.9). SVM and AB achieved the best precision. Our results bring some challenges regarding the autonomy of patients who wish to keep sex information confidential and certainly add greater complexity to the debate about what data should be considered sensitive to the light of the GDPR.

The Dynamism of Transposon Methylation for Plant Development and Stress Adaptation

Plant development processes are regulated by epigenetic alterations that shape nuclear structure, gene expression, and phenotypic plasticity; these alterations can provide the plant with protection from environmental stresses. During plant growth and development, these processes play a significant role in regulating gene expression to remodel chromatin structure. These epigenetic alterations are mainly regulated by transposable elements (TEs) whose abundance in plant genomes results in their interaction with genomes. Thus, TEs are the main source of epigenetic changes and form a substantial part of the plant genome. Furthermore, TEs can be activated under stress conditions, and activated elements cause mutagenic effects and substantial genetic variability. This introduces novel gene functions and structural variation in the insertion sites and primarily contributes to epigenetic modifications. Altogether, these modifications indirectly or directly provide the ability to withstand environmental stresses. In recent years, many studies have shown that TE methylation plays a major role in the evolution of the plant genome through epigenetic process that regulate gene imprinting, thereby upholding genome stability. The induced genetic rearrangements and insertions of mobile genetic elements in regions of active euchromatin contribute to genome alteration, leading to genomic stress. These TE-mediated epigenetic modifications lead to phenotypic diversity, genetic variation, and environmental stress tolerance. Thus, TE methylation is essential for plant evolution and stress adaptation, and TEs hold a relevant military position in the plant genome. High-throughput techniques have greatly advanced the understanding of TE-mediated gene expression and its associations with genome methylation and suggest that controlled mobilization of TEs could be used for crop breeding. However, development application in this area has been limited, and an integrated view of TE function and subsequent processes is lacking. In this review, we explore the enormous diversity and likely functions of the TE repertoire in adaptive evolution and discuss some recent examples of how TEs impact gene expression in plant development and stress adaptation.

K-mer-based machine learning method to classify LTR-retrotransposons in plant genomes

Every day more plant genomes are available in public databases and additional massive sequencing projects (i.e., that aim to sequence thousands of individuals) are formulated and released. Nevertheless, there are not enough automatic tools to analyze this large amount of genomic information. LTR retrotransposons are the most frequent repetitive sequences in plant genomes; however, their detection and classification are commonly performed using semi-automatic and time-consuming programs. Despite the availability of several bioinformatic tools that follow different approaches to detect and classify them, none of these tools can individually obtain accurate results. Here, we used Machine Learning algorithms based on k-mer counts to classify LTR retrotransposons from other genomic sequences and into lineages/families with an F1-Score of 95%, contributing to develop a free-alignment and automatic method to analyze these sequences.

Transfer Learning in Magnetic Resonance Brain Imaging: A Systematic Review

(1) Background: Transfer learning refers to machine learning techniques that focus on acquiring knowledge from related tasks to improve generalization in the tasks of interest. In magnetic resonance imaging (MRI), transfer learning is important for developing strategies that address the variation in MR images from different imaging protocols or scanners. Additionally, transfer learning is beneficial for reutilizing machine learning models that were trained to solve different (but related) tasks to the task of interest. The aim of this review is to identify research directions, gaps in knowledge, applications, and widely used strategies among the transfer learning approaches applied in MR brain imaging; (2) Methods: We performed a systematic literature search for articles that applied transfer learning to MR brain imaging tasks. We screened 433 studies for their relevance, and we categorized and extracted relevant information, including task type, application, availability of labels, and machine learning methods. Furthermore, we closely examined brain MRI-specific transfer learning approaches and other methods that tackled issues relevant to medical imaging, including privacy, unseen target domains, and unlabeled data; (3) Results: We found 129 articles that applied transfer learning to MR brain imaging tasks. The most frequent applications were dementia-related classification tasks and brain tumor segmentation. The majority of articles utilized transfer learning techniques based on convolutional neural networks (CNNs). Only a few approaches utilized clearly brain MRI-specific methodology, and considered privacy issues, unseen target domains, or unlabeled data. We proposed a new categorization to group specific, widely-used approaches such as pretraining and fine-tuning CNNs; (4) Discussion: There is increasing interest in transfer learning for brain MRI. Well-known public datasets have clearly contributed to the popularity of Alzheimer's diagnostics/prognostics and tumor segmentation as applications. Likewise, the availability of pretrained CNNs has promoted their utilization. Finally, the majority of the surveyed studies did not examine in detail the interpretation of their strategies after applying transfer learning, and did not compare their approach with other transfer learning approaches.

InpactorDB: A Classified Lineage-Level Plant LTR Retrotransposon Reference Library for Free-Alignment Methods Based on Machine Learning

Long terminal repeat (LTR) retrotransposons are mobile elements that constitute the major fraction of most plant genomes. The identification and annotation of these elements via bioinformatics approaches represent a major challenge in the era of massive plant genome sequencing. In addition to their involvement in genome size variation, LTR retrotransposons are also associated with the function and structure of different chromosomal regions and can alter the function of coding regions, among others. Several sequence databases of plant LTR retrotransposons are available for public access, such as PGSB and RepetDB, or restricted access such as Repbase. Although these databases are useful to identify LTR-RTs in new genomes by similarity, the elements of these databases are not fully classified to the lineage (also called family) level. Here, we present InpactorDB, a semi-curated dataset composed of 130,439 elements from 195 plant genomes (belonging to 108 plant species) classified to the lineage level. This dataset has been used to train two deep neural networks (i.e., one fully connected and one convolutional) for the rapid classification of these elements. In lineage-level classification approaches, we obtain up to 98% performance, indicated by the F1-score, precision and recall scores.

TIP_finder: An HPC Software to Detect Transposable Element Insertion Polymorphisms in Large Genomic Datasets

Transposable elements (TEs) are non-static genomic units capable of moving indistinctly from one chromosomal location to another. Their insertion polymorphisms may cause beneficial mutations, such as the creation of new gene function, or deleterious in eukaryotes, e.g., different types of cancer in humans. A particular type of TE called LTR-retrotransposons comprises almost 8% of the human genome. Among LTR retrotransposons, human endogenous retroviruses (HERVs) bear structural and functional similarities to retroviruses. Several tools allow the detection of transposon insertion polymorphisms (TIPs) but fail to efficiently analyze large genomes or large datasets. Here, we developed a computational tool, named TIP_finder, able to detect mobile element insertions in very large genomes, through high-performance computing (HPC) and parallel programming, using the inference of discordant read pair analysis. TIP_finder inputs are (i) short pair reads such as those obtained by Illumina, (ii) a chromosome-level reference genome sequence, and (iii) a database of consensus TE sequences. The HPC strategy we propose adds scalability and provides a useful tool to analyze huge genomic datasets in a decent running time. TIP_finder accelerates the detection of transposon insertion polymorphisms (TIPs) by up to 55 times in breast cancer datasets and 46 times in cancer-free datasets compared to the fastest available algorithms. TIP_finder applies a validated strategy to find TIPs, accelerates the process through HPC, and addresses the issues of runtime for large-scale analyses in the post-genomic era. TIP_finder version 1.0 is available at https://github.com/simonorozcoarias/TIP_finder.

TERL: classification of transposable elements by convolutional neural networks

Transposable elements (TEs) are the most represented sequences occurring in eukaryotic genomes. Few methods provide the classification of these sequences into deeper levels, such as superfamily level, which could provide useful and detailed information about these sequences. Most methods that classify TE sequences use handcrafted features such as k-mers and homology-based search, which could be inefficient for classifying non-homologous sequences. Here we propose an approach, called transposable elements pepresentation learner (TERL), that preprocesses and transforms one-dimensional sequences into two-dimensional space data (i.e., image-like data of the sequences) and apply it to deep convolutional neural networks. This classification method tries to learn the best representation of the input data to classify it correctly. We have conducted six experiments to test the performance of TERL against other methods. Our approach obtained macro mean accuracies and F1-score of 96.4% and 85.8% for superfamilies and 95.7% and 91.5% for the order sequences from RepBase, respectively. We have also obtained macro mean accuracies and F1-score of 95.0% and 70.6% for sequences from seven databases into superfamily level and 89.3% and 73.9% for the order level, respectively. We surpassed accuracy, recall and specificity obtained by other methods on the experiment with the classification of order level sequences from seven databases and surpassed by far the time elapsed of any other method for all experiments. Therefore, TERL can learn how to predict any hierarchical level of the TEs classification system and is about 20 times and three orders of magnitude faster than TEclass and PASTEC, respectively https://github.com/muriloHoracio/TERL. Contact:murilocruz@alunos.utfpr.edu.br.

Measuring Performance Metrics of Machine Learning Algorithms for Detecting and Classifying Transposable Elements

Because of the promising results obtained by machine learning (ML) approaches in several fields, every day is more common, the utilization of ML to solve problems in bioinformatics. In genomics, a current issue is to detect and classify transposable elements (TEs) because of the tedious tasks involved in bioinformatics methods. Thus, ML was recently evaluated for TE datasets, demonstrating better results than bioinformatics applications. A crucial step for ML approaches is the selection of metrics that measure the realistic performance of algorithms. Each metric has specific characteristics and measures properties that may be different from the predicted results. Although the most commonly used way to compare measures is by using empirical analysis, a non-result-based methodology has been proposed, called measure invariance properties. These properties are calculated on the basis of whether a given measure changes its value under certain modifications in the confusion matrix, giving comparative parameters independent of the datasets. Measure invariance properties make metrics more or less informative, particularly on unbalanced, monomodal, or multimodal negative class datasets and for real or simulated datasets. Although several studies applied ML to detect and classify TEs, there are no works evaluating performance metrics in TE tasks. Here, we analyzed 26 different metrics utilized in binary, multiclass, and hierarchical classifications, through bibliographic sources, and their invariance properties. Then, we corroborated our findings utilizing freely available TE datasets and commonly used ML algorithms. Based on our analysis, the most suitable metrics for TE tasks must be stable, even using highly unbalanced datasets, multimodal negative class, and training datasets with errors or outliers. Based on these parameters, we conclude that the F1-score and the area under the precision-recall curve are the most informative metrics since they are calculated based on other metrics, providing insight into the development of an ML application.

A comparative study of machine learning and deep learning algorithms to classify cancer types based on microarray gene expression data

Cancer classification is a topic of major interest in medicine since it allows accurate and efficient diagnosis and facilitates a successful outcome in medical treatments. Previous studies have classified human tumors using a large-scale RNA profiling and supervised Machine Learning (ML) algorithms to construct a molecular-based classification of carcinoma cells from breast, bladder, adenocarcinoma, colorectal, gastro esophagus, kidney, liver, lung, ovarian, pancreas, and prostate tumors. These datasets are collectively known as the 11_tumor database, although this database has been used in several works in the ML field, no comparative studies of different algorithms can be found in the literature. On the other hand, advances in both hardware and software technologies have fostered considerable improvements in the precision of solutions that use ML, such as Deep Learning (DL). In this study, we compare the most widely used algorithms in classical ML and DL to classify the tumors described in the 11_tumor database. We obtained tumor identification accuracies between 90.6% (Logistic Regression) and 94.43% (Convolutional Neural Networks) using k-fold cross-validation. Also, we show how a tuning process may or may not significantly improve algorithms' accuracies. Our results demonstrate an efficient and accurate classification method based on gene expression (microarray data) and ML/DL algorithms, which facilitates tumor type prediction in a multi-cancer-type scenario.