GATCGGenerator: New Software for Generation of Quasirandom Nucleotide Sequences

Introduction. DNA is a unique biopolymer that provides storage, transmission and reproduction of genetic information in living organisms. DNA molecules consist of four types of nucleotides containing nitrogenous bases: adenine (A), guanine (G), cytosine (C), thymine (T). Their possible combinations provide nucleotide sequences forming functional genetic elements. In molecular biology and genetics, the basic investigations are carried out on nucleotide sequences of living organisms, but there is an increasing need to create artificial sequences, especially, when solving non-biological tasks (e.g., DNA calculations [1][2], storage in DNA [3], cryptography [4], DNA tags in hydrology [5], etc.).

It is expected that by the end of 2040, the volume of information will reach several yottabytes (10²⁴), which requires its structuring and storage. Both of these processes affect significantly the consumption of energy resources, as well as the production of storage devices and peripheral devices (hard drives, solid-state drives). To store such an amount of information, more than 10⁹ kg of extra pure silicon is required [6], which may not be enough. The solution is seen in using the principles of DNA to work with large-scale amounts of data.

Nucleotide sequences are easily digitized by assigning the corresponding binary codes to individual nucleotides
[7–11] or blocks of nucleotides [12–14]; therefore, text, graphic or multimedia files can be converted into nucleotide sequences [15–18]. Artificial nucleotide sequences can be made manually or generated using special software (DNA generators), depending on the tasks being solved. Some DNA generators were developed as independent applications, others — as part of software packages designed to solve general [19] ^{1, 2, 3, 4, 5} or specific tasks [20]. As a rule, DNA generators are developed on the basis of combinatorial approaches and produce random sequences of a given length of guanine-cytosine (GC) composition. However, such software solutions do not take into account the chemical properties of nucleotides and do not provide obtaining sequences with a certain structure (e.g., without homopolymer sites or long repeating motifs). Therefore, the sequences created by such generators cannot always be reproduced in the laboratory. Moreover, such sequences may be identical to DNA fragments existing in nature, which introduces ambiguity when trying to encode information of a non-biological nature.

The presented work is aimed at creating a generator of nucleotide sequences of a special structure that can be used when encoding text, graphic and other information in DNA molecules.

Materials and Methods. The criteria that should be kept in mind when creating sequences were defined. The need to vary the GC composition, set a certain number of dinucleotides, and exclude homopolymer sites in sequences was taken into account.

A team of authors has developed the GATCGGenerator program in Python 3.6 (Anaconda distribution)⁶. To create a bot⁷ in Telegram, Numpy 1.19 [21] and the Python GATCGGenerator library were used. The solution was provided as SaaS (from “software as a service”), which opened up the possibility of access from different devices and platforms.

Input parameters included the number of sequences, their length, GC composition, and dinucleotide content. The generator excluded repeats with a length of two nucleotides more than four times. The result was presented as a CSV file, which contained the following information: sequence, GC composition, and the number of all nucleotides.

Repeats and homopolymer fragments were stored as a separate list. First, a sequence of four elements was randomly generated (random.choice(nuk), where nuc = 'ACGT'). Then the search for repetitions was performed. If there was at least one item from the list, a new random generation was performed. Next, the GC and NN composition was calculated. If the NN composition did not match the user-defined range, the paired nucleotide was replaced randomly and the GC composition was recalculated. If the sequence matched the input parameters, it was written to a set of sequences.

Below is the operation of the program algorithm.

The requirements for the generated nucleotide sequences were set using Telegram chat. An example of user interaction is shown in Figure 1.

In the framework of the presented work, the functionality of random sequence generators and GATCGGenerator were compared. The differences between the input parameters and the specific nucleotide structures obtained as a result of the program were determined (Table 1).

GATCGGenerator has a broader functionality, it allows the user to specify the number of dinucleotides, create sequences without extended homopolymer sites and repeats that affect the success of the experiment. In existing generators, it is only possible to vary the GC composition.

The program created by the authors of this research generates a given number of quasi-random sequences of nucleotides that do not have homology with natural DNA, but are suitable for molecular biological manipulations.

Research Results. GATCGGenerator allows you to generate specific DNA or RNA sequences from 20 to 5,000 nucleotides long, containing a given number of dinucleotides and not containing homopolymer sites (no more than two identical nucleotides located side by side). More stringent generation conditions can cause a long selection of sequences. As an example, we give a small range of possible content of guanine and cytosine and dinucleotides (e.g., GC composition 45–50 % and NN composition 10–20 %). The operating period of the program for various input data is shown in Table 2.

GATCGGenerator, through more stringent sequence generation conditions, removes the limitations of known DNA generators and creates quasi-random sequences of nucleotides depending on the indicated input parameters. You can specify the required number of sequences, their length, GC composition and dinucleotide content, as well as the nature of the nucleic acid (DNA or RNA). Specifically, sequences created by GATCGGenerator can be used in DNA steganography, applied to protect and transmit information through hiding the message content in the nucleotide
sequence [3].

The proposed software solution (GATCGGenerator) provides obtaining a set of quasi-random sequences of nucleotides depending on user-defined input parameters (type of nucleic acid, sequence length, GC and dinucleotide composition). GATCGGenerator excludes the presence of any nucleotide repeats and homopolymer sites longer than three elements. The generated sequences can be used as service or masking sequences (e.g., in DNA steganography) and are suitable for any non-biological enzymatic manipulations. It is possible to generate numerous artificial nucleotide sequences and use them to create a universal oligotheca suitable for multiple encoding of non-biological data and their long-term storage.

The data presented in Table 3 summarizes the results of the program. For a certain type of nucleic acid (in this case, DNA), the following data is shown: the content of dinucleotides (NN %), the number of generated sequences, their length (nucleotides — nt), and GC composition.

The obtained 30-nucleotide sequences were tested using the Blast tool from NCBI. The absence of 100% homology with known DNA sequences of living organisms was identified. The maximum coincidence was observed for the generated sequences with a length of 25 nucleotides (similarity is about 80 %). This indicates the ability of the GATCGGenerator to generate non-biological nucleotide sequences with high efficiency. It can be assumed that the sequences generated in this way do not have an absolute coincidence with the nucleotide fragments of living organisms.

In this case, special DNA-oligonucleotides of artificial origin containing informative and service parts can be used as a convenient information carrier. Recently, the authors of this work have proposed the use of NYRN-oligonucleotides [14] consisting of:

• internal part (YR)n encoding the encrypted information;
• service (auxiliary) parts S1 and S2 flanking sequence (YR)n (Fig. 2).

The length of the sites (n, k and m) may vary, but the structure of the service parts should provide the successful course of amplification reactions (length more than 18 nt, 40–60 % GC composition, absence of homopolymer sites and repeats). GATCGGenerator allows including NN dinucleotides containing identical paired nucleotides (e.g., AA, GG, CC, TT or UU for RNA), which can increase the specificity of molecular hybridization of nucleic acids.

Discussion and Conclusion. Thus, based on the results of the scientific investigation performed, a software solution (GATCGGenerator) has been proposed, which, in comparison to traditional approaches, assumes more stringent conditions for generating sequences. Due to this feature, the limitations of known DNA generators are removed, and quasi-random sequences of nucleotides are formed depending on the specified input parameters. The obtained 30-nucleotide sequences were studied. The test allowed us to establish the absence of 100 % homology with known DNA sequences of living organisms. The generated sequences with a length of 25 nucleotides coincided as much as possible (by about 80 %).

Note also that in order to hide information in NYRN-oligonucleotides, it is required to mix them with masking DNA. Masking sequences should be similar to sequences of NYRN-oligonucleotides, so that when trying to read hidden information, it would be impossible to recognize them without key sequences. The addressee should know the key sequences — primers to the service sites of NYRN-oligonucleotides. The addressee can decipher the transmitted message by isolating informative nucleotide sequences using a polymerase chain reaction followed by sequencing. A set of NYRN- and masking oligonucleotides can be easily obtained using GATCGGenerator, synthesized, and then stored as an oligotheca. To do this, it is enough to determine the optimal NYRN-oligonucleotides with subsequent filling of the oligotheca. In the future, it is planned to conduct laboratory experiments in order to test the proposed method of storing non-biological information and checking the viability of oligotheca obtained using the generator.