exon-pred-svd

A Rust program that predicts protein-coding regions in a DNA sequence based on Singular Value Decomposition (SVD).

Description

• A Rust program that predicts exons and introns in a DNA sequence.
• The input data must be a DNA sequence in FASTA format.
• The analyzing measure is based on anti-notch infinite impulse response (IIR) filtering and SVD.

Dependencies

peroxide (version 0.31 or more)

• peroxide (https://docs.rs/peroxide/latest/peroxide/)

[dependencies]
peroxide = "0.31"

Installation

You can compile this program by using Cargo. 🦀📦

[e.g]

% cd exon-pred-svd-main
% cargo build --release 

Then, the object file exon-pred-sdv is generated in ./target/release/ directory.

Methods

Numerical representation

To apply a Digital Signal Processing, a DNA sequence composed of 4 types of base represented by { A, T, C, G } letters should be converted into a numerical one. This program assigns a numerical mapping value to each base with the one-dimensional numerical representation method.

This program follows 4 types mapping values assigned as follows (Table. Ⅰ) :

Table. Ⅰ Mapping techniques.

	Adenine (A)	Thymine (T)	Cytosine (C)	Guanine (G)
EIIP [1]	0.1260	0.1335	0.1340	0.0806
Integer [2]	2	0	1	3
Pseudo EIIP [3]	0.1994	0.1933	0.0692	0.0123
Paired numeric [4]	1	1	-1	-1

Filtering stage

The purpose of the filtering stage is to emphasize 3-base periodicity by making peaks. This program employs anti-notch IIR filter with a center angular frequency of 2π/3 as follows [5].

Exons and introns prediction using SVD

Prior to apply the SVD analyzing, the filtered signal is rearranged into a 3 × m matrix A as follows.

✅ NOTE that m must be a multiple of 3, based on [5].

Then, The SVD is applied to the filtered signal as follows [5].

Where A is rearranged 3 × m matrix of filtered signal, U and V are orthogonal square matrices of dimensions 3 × 3 and m × m, respectively. The highest singular value of that SVD reflects the degree of periodicity [5]. It is then processed on a moving frame basis.

Input file format

A DNA sequence in FASTA format. ⚠️ NOTE that a Multi-FASTA format is NOT acceptable, means the input file MUST have just 1 DNA sequence.

See an example input data in ./example/ directory.

Usage

Major arguments :

• -i : Input filename in FASTA format, REQUIRED.
• -o : Output filename, REQUIRED.
• -f : Frame size ( default 81 ). Note that this value MUST be a multiple of 3 ( e.g. 81, 102 and 351 etc. ).
• -m : Mapping method to convert a DNA sequence into a numerical one.

Type -h to see the other available options.

[e.g]

% ./exon-pred-sdv -i sequence_FZ412301.fasta -o output.txt -f 351 -s yes -c no

Output

DNA position \t Degree of periodicity

Additional data

MATLAB source code

It is confirmed that the output result of the program is identical to a MATLAB's whose source code is on ./matlab/ directory. It would help you to see how this Rust program concretely works and it's algorithm in the whole process.

Result data visualization with R

Visit here ! 👈👈👈

References

1. Fickett, James W. "Recognition of protein coding regions in DNA sequences." Nucleic acids research 10.17 (1982).
2. Cosic, Irena. "Macromolecular bioactivity: is it resonant interaction between macromolecules?-theory and applications." IEEE Transactions on Biomedical Engineering 41.12 (1994).
3. Ramachandran, Parameswaran, Wu-Sheng Lu, and Andreas Antoniou. "Optimized numerical mapping scheme for filter-based exon location in DNA using a quasi-Newton algorithm." Proceedings of 2010 IEEE International Symposium on Circuits and Systems. IEEE (2010).
4. Buldyrev, S. V., et al. "Analysis of DNA sequences using methods of statistical physics." Physica A: Statistical Mechanics and its Applications 249.1-4 (1998).
5. Akhtar, Mahmood, Eliathamby Ambikairajah, and Julien Epps. "Detection of period-3 behavior in genomic sequences using singular value decomposition." Proceedings of the IEEE Symposium on Emerging Technologies, 2005.. IEEE (2005).