Bioinformatics or computational biology is a new field of study which is closely related to medicine. It has arrived some 30 years ago and gained a boost of popularity along with the development of the Human Genome Project. It was claimed by the BIS Research Report that, the “global precision medicine market is expected to be $141.7 billion by 2026” [1].

Bioinformatics becomes a very fast-growing field with many scientists involved and large investments present, hence there will be more and more researches related to it. Results of those researches can affect everybody's lives - e.g. invention of new generation medication, based on RNA interference can solve problems of inactivation of mutated genes which could lead to the development of cancer or longevity researches which could suggest methods of increasing human's lifespan. This field of study already implies processing of large amounts of data using specific techniques in order to make it efficient. Biological data is a subject of processing by special algorithms and approaches.

Today genome editing technologies attract a lot of attention because of the fact that they can achieve dramatic impact on our lives: now genomes of different agricultural organisms are being modified in order to add resistance to pathogens or increase fertility. It is theoretically possible to cure diseases by replacing broken genes and endogenous viruses.

Since Barbara McClintock's discovery of the so-called “jumping genes” there was a great interest in these genomic elements. Existence of these transposable elements (transposons) has been demonstrated both in vitro and in vivo. They can be subdivided in two groups by the mechanism of transposition: cut-and-paste (DNA) transposons and copy-and-paste transposons (RNA transposons, retrotransposons). First ones are already being tested as an instrument to deliver genes into a cell's genome, but retrotransposons may become even a more efficient tool for this purpose, because they can produce a lot of copies of themselves. The problem is that the exact mechanism of retrotransposition remains unclear.

This study is devoted to most common human retrotransposons L1 and Alu in order to understand mechanism of transposons' RNA recognition by the reverse transcriptase. The properties of retrotransposons will be evaluated by the machine-learning approach. The study consists of creating a data preprocessing and machine learning pipeline which was applied to biological poorly structured data. Computational part of this study includes annotation of sequences by stem-loop structures and extraction of physical, chemical and geometrical features in order to build a machine-learning model with interpretable predictors.

The goal of the present research is to understand mechanism of retrotransposition, which will be helpful in the design of experimental genome-editing systems. The tasks to achieve the goal include:

- extracting L1 and Alu sequences from human genome;

- annotating sequecnes with stem-loop structures;

- characterizing stem-loop structures with physical, chemical and geometrical properties;

- constructing machine learning models that should be trained to separate stem-loop structures at the 3'-end of human L1 and Alu mobile genetic elements from stem-loops of shuffled sequences.

- extracting the most important properties of stem-loop structures at the 3'-ends of L1 and Alu retrotransposons

1. Big data in bioinformatics

1.1 Emergence of Big Data

The humanity is constantly gathering and storing new information. Evolution of technologies allows creating more and more sophisticated methods to retain that data that results into exponential growth of data stored. For example, “In 2011 1.8 zetabytes of data had been created” [2], there are forecasts that the volume of data will increase by 50 times in 2020 [2].

Capabilities of information processing are developing very fast so some researches, which required power of a mainframe twenty years ago, can be completed using personal computer now. Community is turning towards scalable approaches such as the utilization of cloud resources. This will actually even more increases possibilities of cheap and effective computing. And all these factors help to conduct researches which require a lot of powerful computations including those, related to bioinformatics.

Big Data has great impact on economics: it is present in almost every sphere of human life. It provides a lot of different challenges like building efficient storages and processing pipelines, increasing computational power, inventing new algorithms. Today's market requires a lot of specialists in this field and offers many different opportunities.

1.2 Big Data in Bioinformatics

Bioinformatics is a new interdisciplinary field of science which uses different methods and software for processing and understanding biological data. It transforms sections of human knowledge such as mathematics, physics, computer science, into information processing. This recently developed scientific filed aims towards an integration of heterogeneous data analysis and generation of experimentally testable hypotheses. It has very many different applications such as a disease prediction based on a genome sequencing, understanding evolutionary relations between different species, inventing new medicines, e.g. antibiotics. Bioinformatics has been traditionally used in molecular biology, especially in dealing with data related to genomics.

Since the discovery of the DNA structure by Watson and Crick [3] scientific community was interested in decoding “the code of life” - sequencing a genome of a living organism. It was believed to explain causes of diseases and details of heredity. Solving the task of reading a whole genome of a mammal took almost a half of a century and it happened in 2002 [4]. This delay may be explained if we take into an account several facts:

· Processing raw DNA molecule to obtain its genetic code as a sequence of letters is a very complex task which requires sophisticated tools.

· Full genome of an animal requires large diskspace to be stored, e.g. size of human genome is about 3 000 000 000 base pairs.

Processing of that data is an even more complex task. Genomic sequences are unstructured: coding and non-coding regions are hard to distinguish at a first glance. It took a lot effort and many different researches to isolate genes from other, non-coding parts of genome.

Also, researches revealed that there are some genomic elements which occur in extreme amounts in genome, for example, there is a family of mobile genetic elements called Alu which account for approximately 11% of whole human genome, and there is approximately a million of such sequences in everybody's genome [5].

Hence, we can conclude that processing of such data requires powerful computers, sometimes even cloud-based distributed solutions and sophisticated state-of-art algorithms.

Bioinformatics is a rapidly developing field of science. It has a big potential and it is very popular among people of different professions. Increasingly, big genomic data sets are being used in biotechnology companies, drug firms and medical centers, which also have specific needs. The thoughts shared by these accomplished computational biologists make it clear that biology is becoming a data-related science, and future breakthroughs will depend on strong collaborations between experimental and computational biologists. Biologists will need to adapt to the data-driven nature of the discipline, and the training of future researchers is likely to reflect these changes as well. Aspects of computational biology are integrating into all levels of medicine and health care. And usage of modern computational tools affects further development of new ones because manufacturers have to make tools compatible and operational in different environments. Given that big-data analysis in biology is incredibly difficult, open science is becoming increasingly important.

1.3 Molecular medicine and gene therapy

Molecular medicine is a new trend in the high-tech age. It is a broad field which represents a fusion of biology, bioinformatics, physics, chemistry and medicine. Molecular medicine is pointed towards finding out molecular and genomic issues which cause diseases and developing some kinds of interventions to fix ones. It provides complex treatments that may cure diseases through engineering cells genomes. Even though it has some drawbacks, such as requirement of stable genomic integration methods, it can achieve significant results curing diseases, that previously were perceived as very tough ones.

One of the key features of this approach is that it provides more and more sources of data related to any patient's specific causes. Medical practitioners obtain sophisticated tools to make a factual-based diagnosis. This leads to comprehensive analysis of patient's situations based not only on symptoms, but which also relies on DNA-array based prognosis screening, which results into treatments that correspond best to patient's conditions.

Another feature of the molecular medicine is an idea that some diseases could be treated not only with drugs and other previously used approaches but using the so-called gene therapy. It is a new way to deal with gene malfunctions. Modern biomedical technologies provide powerful tools to interrupt into cell's metabolism and heredity.

This field of study is developing with an impressive speed, for example there were more than 2500 gene therapy clinical studies [6] back in 2016 (see Figure 1).

Figure 1. A) Different vector system contributions to clinical trials that took place in 2016. B) The most popular clinical trials that employed vector system list.

As we see, back in 2016 vast majority of gene delivery systems approved for clinical trials were viral vector-based. Viral vectors are virus-derived tools for delivering something into cells. Around ten years ago the main challenge of gene therapy was not a lack of therapeutic genes, but paucity of efficient gene delivery systems [7]. Viruses are naturally very efficient genetic information “couriers” - this process is highly required for their own replication. Process of engineering them usually implies removing parts of viral genome related to replication. This a good alternative to manual inserting naked DNA directly into cells, because former method may damage or even kill target cells, it is less stable and sustainable.

Viral vector DNA delivery method implies usage of specific modified viruses with needed DNA sequences cloned into viral genome and parts related to replication removed. The former ensures that a virus will be no longer virulent. Vectors are being replicated in laboratory conditions and then passed to the organism (cells) and, depending on a virus type, it may infect up to 100% of the cell population. Thus, this method is proven to be a highly efficient way to deliver genetic material.

However, even vector-based methods have some solid cons. Firstly, they can induce an immune response against vector-encoded proteins from a patient and hence loose efficiency [6], [7].This also results into some therapy-related adverse effects [6]. Another drawback is that large genes integrated into viral vectors dramatically reduce efficiency of those viral packages [8]. Another weak point of such way is sustainability of viral plasmids into eukaryotic cells - only small fraction of viruses inserts the genetic material into host's cells genome, Those are called retroviruses, and HIV is a well known example of the discussed group. Plasmids could be maintained for a long time only in prokaryotes. This fact limits possible carrier options to two viral families (lenti-/retro-). “However, the utility of retro?/lenti- viral vectors is heavily restricted by the size of genes” [8]. There could be problems of another origin as well. Patient could have preexisting cellular immunity against such kind of a virus which would lead to inefficiency of such method [9].

There is another trending approach of editing cell's genome, CRISPR-Cas9 system. CRISPR stands for “Clustered Regularly Interspaced Short Palindromic Repeats”. This mechanism was derived from bacterial adaptive immune system. Functionality of this method is based on ability of CRISPR-Cas9 complex to cut DNA matching to some provided template. This results in target DNA cleavage and reparation by special cell's machinery. It is possible then to supply a reparation template with a desired gene in it, which will result into repairing original cell's genome along with a pattern sequence. (See Figure 2)

Figure 2. Mechanism of DNA reparation after the cleavage by CRISPR-Cas9 complex

One of the problems associated with CRISPR-Cas9 system is that double-breaks in DNA strands reparation is error-prone and it can result in “deletions or insertions leading to loss of function” (See Figure 3) [10]. Another challenge is to increase specificity of this method. Off-target modifications could occur, which may lead to other genes malfunction [10].

Figure 3. Possible outcome of non-homologous end joining caused by CRISPR-Cas9

There is another group of possible genome-editing tools, it is called Zinc-finger proteins (ZNFs). They are an abundant group of proteins and have a wide range of molecular functions [11]. First ZNF was discovered in the late 1980 and it was capable of binding specific sequences of DNA. It is one of the most frequently used DNA binding specific motifs found in eukaryotes. It is possible to exploit this functionality using proteins containing several ZNF active domains each one recognizing specific triplet of DNA letters. Fusing this recognition module with a sequence-independent endonuclease was the first successful strategy to introduce breaks at specific sites of genomic DNA [12].

Even though this technology provides a way to edit the genome, there are some drawback it introduces as well: it is a complex task to construct a complex of several ZNF domains which will recognize desired DNA sequence.

As we see, modern genome-editing technologies have drawbacks and there is no ready-to-go ultimate solutions. But they all provide several alternative methods to modify genomes and hence each one could be substituted by another if it is required to do so. Molecular medicine in general and genome therapy in particular are very important and promising fields of study and more and more new researches should take place.

1.4 DNA and RNA Mobile Genetic Elements

Mobile genetic elements are pieces of DNA which are capable of multiplying and changing their positions inside a genome. There is a strong evidence that they are accumulated in different species and at least some of them are still active [13]. They could be found in almost every genome with very rare exceptions. It is proved with an increase of whole-genome sequencing data available. Transposable elements occupy significant parts of genomes of different organisms (46% of human, 40% of rat, 85% of corn) [13].

Human genome harbors significant quantities of mobile genetic elements, some of which are still active and sometimes they can act as a “tool” of evolution - modify existing genes by making insertions into them, change expression of genes by breaking promoters or “jumping” nearby.

Generally speaking, there are two distinct types of transposons which are identified on the base of their mechanism of displacement: DNA and RNA (retrotransposons). Key difference between them is a mechanism of transposition: former ones move by cut-and-paste method and first is using copy-and-paste method (See Figure 4). Both classes are subdivided into families and subfamilies by sequences similarity, structure or transposition mechanisms.

Figure 4. Transposon classes distinguished by transposition intermediate steps

Retrotransposons use RNA intermediate in the process of transposition - e.g. they are transcribed into mRNA from DNA and then retrotranscribed back into DNA. The last step is an insertion of newly built sequence into organism's genome.

On the other hand, DNA transposons are not being retrotranscribed, they are being moved, not copied. This strongly affects their replication mechanism - they are being copied through “indirect mechanisms that rely on the host machinery” [14].

Transposable elements were considered a “junk” DNA for a long time, now they are believed to play a significant role in cell's life. They can take part in gene expression regulation using their own promoters, affect alternative splicing (see Fig. 5), even trigger chromosome rearrangements [15]. In some cases they play a significant role in diseases. Transposon activity can lead to interruption of gene sequence by a jumping gene itself that could possibly lead to gene's malfunction. It also can result into deletions or chromosomal rearrangements [15].

Figure 5. Simplified view of the main consequences of transposon affecting splicing of a gene.

This study considers RNA transposable genetic elements, because they are the only transposons that are still capable of displacement into human genome (there are experiments that prove this by showing de novo insertions) [16]. There are three families of human RNA mobile genetic elements that are currently active in vivo - long interspersed nuclear elements class 1 - L1, Alu, which was firstly associated with action of the bacterium Arthrobacter luteus (Alu) restriction endonuclease (gene-cutting protein), hence the name, and SVA (SINE-VNTR-Alus), which are somewhat alike with endogenous retroviruses [17]. Former transposon family is out of scope of this study.

For some species it was previously experimentally shown that replication of transposable elements highly depends on RNA secondary structures - stem-loops - located at the end of transposon sequence [18]. It also was shown in silico that stem-loop structures are present in almost all Alu and L1 mobile genetic elements in human genome [19], thus, these structures may be crucial for retrotransposition of even wider retrotransposon range.

1.5 RNA secondary structures

Nucleic acids secondary structures are forms in which RNA or DNA could fold itself by base-pair interactions. RNA secondary structures differ from DNA structures because DNA exists mostly as double helix in contrast to RNA, which is mostly represented as a single strand. Also, RNA is more likely to form some complex and intricate base-pairing interactions because of its increased ability to form hydrogen bonds due to the extra hydroxyl group located in the ribose sugar.

Stem-loop secondary structures, also known as hairpins or hairpin loops are base-pairing patterns that commonly occur in single-stranded RNA (see Fig. 6). These structures are common in all genomes, there are researches that related to discovery of non-random stem-loop patterns in prokaryotic [20] and eukaryotic genomes.

Figure 6. RNA Stem-loop structure example.

This study includes transposon sequence stem-loop annotation in order to obtain more interpretable feature extraction.

1.6 Possible applications of transposons

Transposons are subjects of many researches. Scientists tend to have an interest in studying this topic due to possible outcomes: biological mechanisms that are capable of changing cell's genome in vivo are extremely powerful - and may be harnessed as tools to engineer cell's genomes [21].

Newly discovered mobile genetic elements may be used in order to achieve different results. We can consider Sleeping Beauty DNA transposable element as an example: it originates from some fish species and has some interesting features. Firstly, it was reconstructed from an ancient and currently nonactive transposon found in the fish genome [8]. Secondly, it was modified to increase its efficiency by 100 times [8]. Now it is often used in genomic therapy as a non-viral tool for delivering and inserting DNA into target cells.

Retrotransposons naturally can take part in some genome-rearrangement processes in living organisms. For example, it is known, that LINE-1 (Long Interspersed Nucleotide Elements) can promote gene duplication [22]. Generally speaking, retrotransposons play significant role in cells, they may provide substrate for DNA reparation systems during nonallelic homologous recombination [22], alter gene regulation, take part into cell regulatory pathways. It is known that up to 63% “of primate-specific regulatory sequences are derived from transposable elements” [23].

Such elements should not be ignored by science because of the abundance of possible applications. RNA transposable elements are not used as a genome editing tool yet, but it is clear that understanding the way they work may give us a possibility to exploit them and widely affect genomes of living organisms.

Summary