Open access peer-reviewed chapter

New Trends in Artificial Intelligence: Applications of Particle Swarm Optimization in Biomedical Problems

Written By

Aman Chandra Kaushik, Shiv Bharadwaj, Ajay Kumar, Avinash Dhar and Dongqing Wei

Submitted: 25 November 2017 Reviewed: 09 January 2018 Published: 13 February 2018

DOI: 10.5772/intechopen.73606

From the Edited Volume

Intelligent System

Edited by Chatchawal Wongchoosuk

Chapter metrics overview

1,452 Chapter Downloads

View Full Metrics

Abstract

Optimization is a process to discover the most effective element or solution from a set of all possible resources or solutions. Currently, there are various biological problems such as extending from biomolecule structure prediction to drug discovery that can be elevated by opting standard protocol for optimization. Particle swarm optimization (PSO) process, purposed by Dr. Eberhart and Dr. Kennedy in 1995, is solely based on population stochastic optimization technique. This method was designed by the researchers after inspired by social behavior of flocking bird or schooling fishes. This method shares numerous resemblances with the evolutionary computation procedures such as genetic algorithms (GA). Since, PSO algorithms is easy process to subject with minor adjustment of a few restrictions, it has gained more attention or advantages over other population based algorithms. Hence, PSO algorithms is widely used in various research fields like ranging from artificial neural network training to other areas where GA can be used in the system.

Keywords

  • PSO
  • DNA
  • medicines
  • biological complexity
  • biological problem

1. Introduction

Particle swarm optimization (PSO) is defined as computational procedure to recruit or select the most effective element from a collection of accessible alternatives [1]. Associated optimization drawback either deals with increment or minimization in the true operation for simplest state of affairs while constantly screening the input elements at intervals associated with allowed set of accessible alternatives [2].

In 1995, PSO was purposed by Dr. Russell Eberhart and Dr. James Kennedy after inspired by flocking and schooling patterns of birds and fish, respectively. Originally, PSO was established with an intent to design laptop codes based on simulations of bird flocking around food sources. However, developers later realized and established PSO algorithm utility in resolving optimization problems [3].

Albeit, PSO sounds complicated and complex but it is very simple and straightforward algorithm. According to this algorithm, a set of variables will tend acquire a value close to the element with highest value of the target at any given interval or moment. For instance, imagine a flock of birds hovering over a region where they can sense hidden supply of food. The one close to food will chirp loudest while other birds will follow the path toward him. Additionally, even if one bird circling opposite approaches nearer to food than primary bird, then this bird chirp louder that makes other birds to follow him. Likewise, this process continues until all the birds landed on food. Hence, PSO was recommended as an easy and simple algorithm for implementation [46].

Advertisement

2. Basic PSO algorithm

In PSO algorithmic program, every individual is termed as “particle or element” and subjected to acceleration in an especially multidimensional space or region that represents the probability area. Also, particles possess the ability to store information as memory and thus, remember their respective previous position in that region. Moreover, no restrictions are applicable to the particles and they equally share the purpose in the postulated space while retaining their uniqueness conserved. So, every element displacement movement is consequences of associated primary random speed and arbitrary weighted effects that includes distinctiveness and nature of the particles, tendency of particles to come back on their respective most suitable previous position as well as particles trick toward the neighborhood’s best previous positions [7]. Basic PSO algorithmic program can be classified further into two versions as continuous PSO algorithm and binary PSO algorithm.

Advertisement

3. Continuous PSO algorithm

This type of algorithmic program finds the feasible area for elements is a real-valued multidimensional region, and develops the location of each particle in such a way that search region can be victimization by use of subsequent Eqs. (1) and Eq. (2) [8]:

V id t + 1 = ω . ν id t + C 1 . ψ 1 . p id t x id t + C 2 . ψ 2 . p gd t x id t E1
X id t + 1 = x id t + ν id t + 1 E2

where ν id t stands for dimension “d” of ith element velocity in repetition t, x id t stands for dimension “d” of ith element position in repetition t, c 1 , c 2 stands for constant weight factors, p i stands for best position attained so long by particle i, p g stands for best position found by neighbors of particle, ψ 1 , ψ 2 stands for random aspects in the [0, 1] interval, ω stands for inertia weight of particle.

Herein, most suitable position for particle in the feasible region discovered by neighbor’s particle (Pg) is depend on the category of nominated neighborhood. However, in basic PSO algorithm, generally a global (gbest) or local (lbest) neighborhood is employed for finding a region for particle in the search region. For the implementation of global neighborhood, all neighbor particles in feasible region are considered for calculating Pg. But only certain or selected number of particles composing the neighborhood among the whole set of population in the search area are considered in case of local neighborhood. It is important to mentioned that a given particles does not change its local neighborhood during the iteration of algorithm.

Also, during implementation of algorithmic program, a limit ν max is forced on ν id t to certify junction. Importantly, the respective values will be remained unchanged in selected interval x id max , + x id max while maximum value of x id max is kept fixed for each element location. Moreover, particles large inertia weight ω is an advantage for the global search while insignificant inertia mass supports local exploration in basic PSO algorithmic program. In some cases when inertia of particle is employed, local search drastically and linearly decreases starting from the initial value during algorithm iteration. In such cases, an alternative formulation of Eq. (1) can be done with replacement of the velocity constraint ν max by a constriction coefficient. In conclusion, PSO algorithm requires modification of certain parameters: the individual and Inertia weights c 1 c 2 and the inertia factor ω . Hence, both theoretical and empirical available studies can be used to select the most suitable scores for the particles [8, 9].

Advertisement

4. Binary PSO algorithm

A binary PSO algorithm program have been additionally developed. This version of PSO gained less attention compare to previous versions of the program. Although, particle position is not a true value in binary (0 or 1), but logistic function of element velocity is generally employed to calculate feasible region for the particle location. In other words, element location is arbitrary in the search region generated by victimization of distribution. Hence, Eq. (3) can be used to study the updated particle position in the feasible distribution region [10]:

x id t + 1 = 1 if ψ 3 < 1 1 + e ν id t + 1 0 otherwise E3

where ν id t stands for dimension “d” of ith element velocity in repetition t, x id t + 1 stands for dimension “d” of element position in repetition (t + i), ψ 3 stands for arbitrary feature in [0, 1] time.

In conclusion, binary PSO algorithm program still conduct the random search for the particle position in feasible region (place in each direction where there is a probability of finding 0.5 to become either 0 or 1) without the influence of individual and social influences. The selection of parameters on binary version of PSO algorithm has not been widely studied and hence, various problems are still demands to be solved. However, some modifications on the binary algorithmic equations have been proposed in the form of quantum approach. Additionally, recent studies have been focused on the classification issues. However, recently evolutionary programming (EP) have been successfully used to solve the various numerical and combinatorial optimization problems [11].

Advertisement

5. Multiple sequence alignment

Multiple alignment of macromolecule sequences serves a critical role in various applications such as phylogenetic tree estimation and secondary structure prediction. Albeit, different approaches have been documented on the sequence comparison that differ from info searches to secondary structure prediction of the macromolecule. However, this approach involves comparison between two or more sequences by aligning them to predict the result of evolution across an entire macromolecule set. However, this process is not easy a task and under such conditions, a typical heuristic is followed for the alignment of multiple sequences in such a way to enhanced SP score. Hence, aspects of algorithms for alignment of multiple sequences have been an awfully dynamic analysis field [12].

Advertisement

6. The sequence alignment problem

In bioinformatics, foremost vital information stored in the form of codes on biological sequences sets that include deoxyribonucleic acid (DNA) sequences and macromolecule sequences. For instance, a DNA sequence is composed of four string nucleotides represented by symbols, i.e., A, C, G, T while a macromolecule sequence such as protein may vary up to 20 amino acid symbols set. Interestingly during evolution, there may occurred insertion, deletion or changes in the segments of biological sequences. Hence, to spotlight the similarities among various selected sequences, generally suitable way is to add distance between the nucleotides to obtain better range of sequence similar regions. The similarity of aligned sequences can be calculated by using rating operator that relies on the matrix and allocates score to each type of codes (mutation based probability). For instance, foremost ordinarily employed matrices for proteins is percept accepted mutation (PAM) and blocks substitution matrix (BLOSUM) [34]. Herein, addition of small distances is required to circumvent the insertion of associated set of amino acids. The method of discovering an associate optimum match among different orders is commonly termed as sequence alignment [12].

Advertisement

7. PSO to the rescue

The sequence alignment drawback may be thought-about as associate optimization drawback within which target has to maximize the rating operate. Thus, PSO algorithmic was tailored to employ for biological sequences. Within tailored PSO algorithm, an element signifies a sequence alignment. Since, key protocol of PSO algorithm depends on particle movements toward lead particle, appropriate operators are projected to employ this mechanism. The overall algorithmic program is described as [12];

Produce a group of initial elements.

Confirm the lead element gbest.

Replication till conclusion condition is fulfilled, i.e.,

  1. live displacement between gbest and each element.

  2. movement each element toward gbest.

  3. confirmation the leader particle.

The end conditions for algorithmic program is the maximum number of repetitions, or several repetitions after that most suitable score does not show any further improvement. Thus, contained thought of PSO algorithm is a set of elements arbitrarily distributed over a search region that are progressively moving to a location where swarm discovered a result that cannot be improved any longer [12].

The projected algorithmic, termed as PSOMSA, was enforced to check its performance. The quantity of particles decided to take under consideration; length of sequences and number of sequences considered for alignment. Hence, to check algorithm program, different eight sets of macromolecules completely with different dimensions and identity percentages sequence were hand-picked from arrangements info predicted by BALiBASE. Finally, one macromolecule set was hand-picked from every length class (short, medium and large), and one from every of distinctiveness proportion (less than 25%, between 25% and 40% and greater that 35%) [12]. The protein families used in this program are shown in the Table 1.

S. No. Name Identity Length PDB ID
1. b-galactosidase 20–40 Long 1AC5
2. Aldehyde dehydrogenase >35 Long 1AD3
3. Repressor <25 Short 1R69
4. Elastase >35 Medium 1EZM
5. Cytochrome p450 <25 Long 1CPT
6. Hiv-1 protease >35 Short 1FMB
7. Cardiotoxin 20–40 Short 1TGX
8. Alpha-trichosanthin 20–40 Medium 1MRJ

Table 1.

The protein families with their identity, length and PDB ID; this data retrieved from http://dsp.jpl.nasa.gov/.

These macromolecule families were antecedent aligned victimization by the well-known algorithmic program Clustal X. It was observed that PSOMSA algorithm program have superior performance in comparison to Clustal X, particularly once the information has smaller sequences and shorter length. But once information features such as longer length results are comparable, still there are several enhancements that can be done in PSOMSA to obtained improved results. Besides, new fitness functions supported totally dissimilar rating strategies are feasible for easy developments [12].

Advertisement

8. Tumor classification using hybrid PSO and Tabu search approach

In present scenario, high-density DNA microarrays are the foremost advanced tools employed in genomic studies. The advancement in microarray technology permits to study simultaneous expression stages of many genes at the same time. Recent studies have documented the use of microarrays in tumor classification. Herein, factor choice played an important role in the factor expression-based tumor classification systems. Microarray experiments produce immense gene expression datasets, however, comparably in less time than conventional techniques. Also, most of genes monitored in microarray technology could also be orthogonal for study and probably constrain the forecast performance by classification rule through involvement of related genes. This drawback is factor choice and assortment of inequitable genes is crucial to raise the accuracy, and to reduce the process complexness and value. By choosing relevant genes, typical classification methods can be useful for microarray generated information. Moreover, factor choice could also highlight those relevant genes and might altered the biology to achieve vital perception into genetic nature of disorder, as well as possible mechanisms to solve it [13].

PSO at the server of Tabu Search (TS) have been operated to design a hybrid algorithmic program that aimed at factor based choice for tumor classification, termed as HPSOTS. The essential steps followed during the execution of this hybrid algorithmic program are as follow [13];

  1. Step 1. To arbitrarily adjust every initial binary strings IND in HPSOTS using an appropriate set of elements, followed by evaluation of fitness function of everyone in IND. Further, IND then played the role of strings in binary bits equivalent to individual gene.

  2. Step 2. To generate and assess neighbors of ninetieth individual in IND that is consistent with info reflecting mechanism of PSO.

  3. Step 3. To adopt fresh discrete elements from explored neighborhood consistent with the objective criterion, tabu conditions and update the set of IND elements.

  4. Step 4. To boost additional power for HPSOTS to overleap native targets, 100% different particles in IND are forced to fly randomly rather succeeding two best elements; assess fitness operate of those tenth of elements.

  5. Step 5. If the simplest object operates for generation fulfills the top conditions, coaching will be stopped along with output results, otherwise, it will visit the second step to renew population.

The selection of factors that are extremely indicative of tissue classification may be included with key phase for emerging triple-crown gene expression-based information study program. A hybrid PSO and TS (HPSOTS) method for factor choice might be employed for tumor classification. The incorporation of TS as neighborhood enhancement practice allows HPSOTS algorithmic program to overleap with local optima and results into fitting performance.

Advertisement

9. Operon prediction

In prokaryotic organisms, they contained one or additional repeated genes, at equivalent DNA that are interlinked for biological function regulations are termed as Operons. These set of genes are encoded into a single-strand of RNA sequence. This concurrent transcription of more than one gene was concluded to contribute equally in biological roles and straight effect the regulation of one another. Hence, DNA prediction are employed to deduce function of theoretical proteins present encoded by different genes on the same sequence. A widely recognized example is lac operon in Escherichia coli. Herein, DNA comprises three uninterrupted structural genes; lacZ, lacY and lacA, that contributed to equivalent promoter and eradicator [14, 15].

Operons in microorganism genomes also comprise information for drug style and deciding macromolecule roles. For example, gram-positive staphylococci bacteria may be a human infective agent that is answerable for community-acquired and healthcare facility infections. Thus, DNA prediction on these bacteria will facilitate drug target identification and can be used for the designing a potential antibiotics. However, available data on operons is less and experimental approaches for envisaging the operons likely to be tough for implementation. In order to achieve higher insight, quantity and association of operons in microbial genome need to be examined in greater details. In this regard, well understanding of rules for transcription is crucial because it enable researchers to precisely envisage the operons association with genome of the microorganism.

Many scientists have anticipated properties that can be applied precisely to forecast the operons. These assets can be further divided into five subsequent groups as: intergenic distance, preserved factor clusters, sequence order and experimental proof. In each of same classes, it is crucial to discover the promoter and eradicator on DNA margins to spot biologically most characteristic properties. The only associated important forecast property is to watch whether gap between factor sets at intervals in DNA is smaller than gap among set of genes at edges of transcription units (TUB pairs). The gap stuff yields excellent DNA prediction results [15].

Many algorithms program are anticipated to suitably counter the sensitivity and specificity in DNA forecast. Jacob et al. projected associate algorithmic program target-hunting by symbolic logic. Symbolic logic does not trust advanced mathematical methods to compute fitness standards of a body. GA employed intergenic gaps, biochemical pathways, cluster of orthologous teams and microarray encoded information to envisage set of genes in the genome. Jacob et al. had also purposed support vector machine (SVM) algorithmic to envisage the operons. This technique employed biological assets as SVM input vectors while splits these factor gathered into operon pairs as well as non-operon pairs [16]. Also, a comparison analysis has been conducted on extra predictors genome-specific, DVDA, FGENESB, ODB, OFS, OPERON, JPOP, VIMSS, UNIPOP and genome-wide DNA forecast in Staphylococcus aureus besides preceding strategies [15].

BPSO may be a rehabilitated algorithmic program for DNA prediction. To validate the practicability of tactic, index chance on individual assets of E. coli order as fitness score of each aspect within element is considered. The genome of three microorganisms; Bacilli sp., Pseudomonas aeruginosa PA01 and Staphylococci aureus were hand-picked as benchmark genomes as celebrated DNA structure. In an exceedingly opening, a constraint was applied within the strand to initialize the basis of intergenic distance property. To pick simplest and doable amalgamation of properties, an idea of feature choice was applied for DNA prediction. The five options were investigated, i.e., intergenic distance, metabolic pathways, COG, factor length magnitude relation as well as DNA length. Hence, based on experimental data; intergenic gaps, biochemical pathways and factor length magnitude relation are nominated screened by feature choice method to compute fitness score for every factor of the exceedingly element.

In BPSO, every element epitomizes a contender resolution to matter while swarm comprises “N” number of elements moving in D-dimension for exploration region till conditions for restrictions are fulfilled. The associated inertia weight with score of one is employed at every generation. The gbest value will be attained when most range of 100 repetitions have been deceased and then allowed us to discuss the steps of algorithm program one by one as follow;

  1. Step 1: Every particle is initialized with support of factor strand and arbitrary threshold score (0 to 600 bp).

  2. Step 2: For every factor, pair score is measured and supported its assets.

  3. Step 3: Fitness value for considered DNA is measured using equation.

  4. Step 4: Fitness value for every element is measured using equation.

  5. Step 5: Every element is rationalized and supported by PSO modified formula while an enquiry on pbest and gbest for the population is recorded.

  6. Step 6: Steps three as well as four will be recurrent till elements fulfilled stopping criteria.

BPSO was employed to forecast operons supported by intergenic gaps, biochemical pathway and factor dimension magnitude relation assets. The experimental data showed that purposed technique not only solely increases the accurateness for DNA forecast by 3 order information on tested sets but it also further reduces execution time required for calculation process [15].

The principle reason to improve the work, that allowed practicing suitable resolution for underneath set, is modern-day constraints supplied to us. Present days, scientist with emblem in new modern days offered several answers to non-linear and linear improvement troubles without any doubt. Affiliation with arithmetic in nursing improvement downside includes a fitness perform representation. This depend on group with cutting-edge constraints that represents the answer to this relied extensive varied area. Typically, normal development strategies hold square degree focused over evaluation of modern day for primary derivatives. So, we can find out superior solution on some well-equipped grounds in nature.

Preliminary goal is to find out optima for several unusual and tough development surfaces. Presently, many amended algorithms which are free from spin country-of-the-art have been purposed and applied. The improvement in algorithms is associated with nursing smart and searching for disadvantage; wherever one or a whole set of modern-day seller’s rectangular diploma used to train session targets on a study of landscape and representing unnatural ground for development of drawback.

Later in cutting-edge of twentieth century, Netherlands pioneered an opportunity insight on natural method for trying to find algorithms, and originated a method to this point open-ended disadvantage to non-linear development issues. Stimulated through herbal variations in modern-day organic species, Netherlands resonated Darwinian idea by his maximum up to date and popular gadget, currently termed as genetic algorithms (GA) [17]. Netherlands and his co-workers together with Goldberg and DeJong, generalized GA concept. In this technique, organic crossovers and modern chromosomal mutations may be realized to elaborate the identical antique modern day, the answers over consecutive iterations. In the middle 1990s, Eberhart and Kennedy postulated a change preference to the advanced non-linear improvement downside with useful resource of latest modern manner emulating the collective behavior of present day chicken flocks and particles. The birds approach, cutting-edge craig painter and socio-cognition known as their product of PSO. While, Worth and Storn took perilous decision to replace the critically overlapping and mutation in GA through unique operators. Thus, they purposed the precise differential operator to address the problem. They projected an opportunity system supported this operator, and referred to as it differential evolution.

Every algorithm does not need any gradient info while extremely-modern-day carry out the optimization through certainly primitive mathematical operators and square degree methods. They may be enforced in any programming language rather truly and minimization of parameters for standardization. Approach normal performance does no longer visit pot severely with growth cutting-edge search location dimensions. Those problems possibly have cutting-edge-day algorithms inside the region holding contemporary tool intelligence and informatics [17].

The use of drastically studied thermodynamic version, a detailed research on modern-day several requirements for designing sequences that intended to adopt a goal as secondary form [18]. The format contemporary DNA and RNA sequences are crucial for loads endeavors, from DNA nanotechnology, to PCR-based totally simply software as well as DNA hybridization arrays. Outcomes inside literature rely upon modern day layout standards tailored specific to the necessities of modern-day software. Generally, strategies used are from extremely-present day collections of symmetry minimization and minimum free-strength pride usually positioned into effect awful format, and may be bolstered by introducing a powerful layout trouble [18]. The superior layout techniques need to explicitly place into effect of each format paradigm (optimize affinity for the purpose shape) and a horrific format paradigm (optimize speciality for the aim form). Also, it has been observed that designing of thermodynamic balance does not determine folding kinetics, emphasizing the opportunity for extending layout requirements to goal kinetic in modern-day power landscape [18].

Advertisement

10. Category selection and classification

We began out our evaluation via exploring the IEEE Xplore record for credentials similar to the hunt word “particle swarm optimization.” Besides, dividing PSO programs into regions is hard. So, as we stated inside the preceding phase, element mechanized approach. This comes to be completed through latest programs. After manually aside from determiners, adjectives, etc., this process generated over 4000 exclusive terms. Following, we used vital detail phrases to describe similarity relationship among the used articles to state a similarity among main terms. The collected information has been manually examined for importance of problem in modern-day packages. A total of 928 main phrases remained after the filtering method. Each paper has grown to be assigned a vector with 928 factors. Every detail showed, how regularly modern key time seemed to choose out and summaries the ultra-modern corresponding paper. Each detail represented how ultra-current corresponding key time seemed to call abstract and modern-day papers. Further, each key time period emerges as assigned a vector with 1100 factors. Further, we stated that similarity among key terms because of reality of scalar results produced are the equivalent vectors. We can hire graphical illustrating applications to gather graphic depictions of present day PSO software papers, PSO software regions and their associations. Those similarities relationships result into respective graph structures in the paper and main term “areas,” in which every articles links to a node and nodes are connected via ends, likeness among the corresponding papers/terms in above a few prefixed thresholds. For that reason, we used modern open-deliver sketch software application Neato (http://www.Graphviz.Org). On this application software, weigh the ultra-modern rims in a graph and results are interpreted as attractive forces among nodes. Starting from a random placement, this tool iteratively adjusts the present-day nodes in and try to minimize the energy in the device. Attractive forces are balanced thru an essential repulsive pressure among nodes to prevent the graph from collapsing to a single detail. Also, zooming on regions in which excessive density ultra-present-day hyperlinks are present, it is easy to gather semi quantitative evidence from ultra-modern which can be critical for software regions in PSO [19].

11. PSO applications

Parkinson’s disease: Natural, clinical and pharmaceutical programs also getting introduced as from total publication in IEEE explore database. About 4.3% covers software program papers in these subjects. Programs includes that of human tremor during Parkinson’s disease for evaluation by modern-day inferences. These includes commonly applied programs; for instance, contemporary gene regulatory networks, Parkinson’s ailment, phylogenetic tree reconstruction, maximum cancers elegance [20], human motion biomechanics optimization, survival prediction, gene clustering, DNA motif detection, protein form prediction and docking, identification of issue binding web sites in DNA, evaluation modern-day thoughts magneto encephalography information, biomarker choice drug format, radiotherapy making plans, biometrics, electroencephalogram assessment and RNA secondary shape willpower [19].

Breast cancer sample: Proposed state-of-the-art information on mining approach that is primarily based on vital idea of contemporary day. The identical antique PSO especially discrete PSO. This segment aimed in route modern that include developing a totally unique PSO, wherein every particle turns out to be coded in terrific integer numbers and has a likely device form. Primarily based on received effects, proposed DPSO can detect the sensitivity up to 100% with 98.71% accuracy and 98.21% specificity. While in the assessment with previous research, the proposed hybrid approach suggested development in each accuracy and robustness. Consistent with excessive modern-day, the proposed DPSO information mining set modern day pointers may be used because the reference for making preference in clinic can be selected as reference by the researchers [21].

Image processing: Consequences of modern day optimizing the normalized mutual data for similarity metric, numerous evolutionary techniques have been employed for comparison. Intensely, hybrid particle swarm method produces better accurate registrations than that of evolutionary techniques in hundreds modern day instances, with comparable convergence. The results displayed that particle swarm techniques as component current evolutionary techniques and close by techniques, are beneficial in photograph registration as well emphasize on the need of hybrid techniques for tough registration problems [22].

Optimization is geared inside the path contemporary to identifying viable method for problems underneath a given set of current situations. PSO is computational approach for optimizing a problem by use of iterative development present day candidate that answer nearly approximately a given set contemporary constraints. Candidate solutions featured with particles and PSO used to optimize a problem by aid contemporary-day transferring them to search for regular place with an easy mathematical relation, primarily based on their pace and function. Every particle is guided in a direction by cutting-edge day stated feature for searching out vicinity. However, motion is usually recommended with useful resource current of said features by the community. The regarded functions that seeking out vicinity gets up to date as higher positions are decided with beneficial and useful resource from present day numerous particles.

This expectedly actions of swarm in the direction modern can be used for possible answer. The inter- and intra-base pairing interactions between molecules that result into secondary forms of nucleic acids. Modern-day, truth of pattern today’s base-pairing determines the overall form of molecules and secondary form influences the layout state-of-the-art by employing nucleotide structures. Moreover, stem-loop also referred as hairpin or hairpin loop is an intra-molecular base pairing sample taking place in the single-stranded DNA or in the RNA. Such base pairing occurred at the identical time of regions identical to the strand base-pair to form a double helix and terminating in an unpaired loop. This serves as a template for hundreds RNA secondary systems. The stem loops have a function in law extremely-contemporary replication, translation and transcription in both prokaryotic and eukaryotic systems.

Recently it was documented that in bacterial macromolecules such as RNA stem loop complements with fantastic expression of non-expressible genes in its genome. Whereas, RNA stem-loop is perception to overwhelm prolonged variety of communications at the initiation site for translation and gene transcription for the specific mRNA sets. Initiation sites for translational process holds attraction toward the ribosome binding sites and has prolonged-installation one after the other. Recently, HIV type 1 virus was observed with duplex establishment because of nucleotides in stem loop that play a critical role in viral contamination and macromolecular complex of the entire genome. In addition, gag gene which comprises stem loop-1 and stem loop-2 were concluded to serve as number one encapsidation indicators in the bovine leukemia virus (blv).

This state-of-the-art channelized the law that present day stem loop areas may be useful in blunting the virulence of modern-day that includes many sickness inflicting microorganisms. Accurate prediction of modern-day DNA/RNA stems will contribute to channelize the regulation. The palindromic sequences shape in backbone of current-day bacterial interspersed mosaic elements (bimes). Typically, in prokaryotic genomes, non-coding areas are brief while characteristic gene expression regulating elements referred to ultra-modern reality and precept elements. However, presence of trendy several household’s modern repetitive factors in one’s intergenic areas; which can be shorter than insertion sequences (IS) and normally lack protein coding functionality as well as their functionality, stays eluding the researchers. Bacterial interspersed mosaic elements (BIMES) are new operons or among the co-transcribed areas. Approximately, 500 BIMES are identified to be scattered over the genome modern-day E. coli.

The BIMES are composed of mosaic aggregate with numerous preserved motifs: palindromic unit (pu) that is also termed as repetitive extragenic palindromic sequences. It contains seven pu adjoining sequences, and additional sectioned into 3 variations divergent set. Numerous combos have concluded 2 households in the BIMES ‘of bacteria E. coli. BIMES -1 which covers sequences such as pu’s (y and z’), at the same time as BIMES-2 consists of 12 pu’s (y and z’) that is probably through repetition modern-day assembly present day-day subjects. Those BIMES emerges to have several set of modern abilities which include mRNA maintenance, transcription, translational and at genomic levels. Furthermore, function of BIMES within beneficial enterprise has been employed in the construction of ultra-modern bacterial chromosome. Consequently, precise forecast BIMES areas can bring about operative regulation in latest bacterial gene expression. Hence, this intention based totally simplifies the drug development and improvement. PSO proposed through way modern day purposed in 1995 by Kenedy and Eberheart, is an improvement to the set of current pointers. It is higher-level population derived procedure of optimization set with current recommendations that devices stimulated with the useful resource cutting-edge-day evolution like reproduction, mutation, recombination and hundreds current others. In interval of ultra-modern time, estimation of competencies in modern-day set suggestions has been utilized and employed at each network on international scale to benefit structural progression. Whilst, neighborhood PSO set present day-day policies can be beneficial in keeping modern early union and improvements in prediction skills at the same time as the global set modern-day suggestions can also use to get the result in speedy convergence with accurate outcomes. The principle motives of cutting-edge is to format a device which can works at PSO set of ultra-contemporary suggestions and prediction as it is required for DNA/RNA and many other modern-day preserved motifs. A number of other algorithms are available, but they are not quite efficient and accurate to solve biological problems [2325].

12. Conclusion

PSO best suited resolution to a haul underneath a given set of biological constraints such as analysis of the tremor in human for Parkinson’s diagnosis, logical thinking of factor restrictive networks, biological pathway designing, identification of cancer category, detection of DNA motif, factor agglomeration, selection of biomarker, medicine style, irradiation designing, brain magnetoencephalography analysis, polymer secondary structure prediction, EEG study, biometry.

Acknowledgments

This work is supported by the Key Research Area Grant 2016YFA0501703 from the Ministry of Science and Technology of China, State Key Lab on Microbial Metabolism, and Joint Research Funds for Medical and Engineering & Scientific Research at Shanghai Jiao Tong University. The simulations in this work were supported by the Center for High Performance Computing, Shanghai Jiao Tong University.

References

  1. 1. Holland JH. Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence. Cambridge: MIT Press; 1992
  2. 2. Eberhart RC, Shi Y, Kennedy J. Swarm Intelligence (The Morgan Kaufmann Series in Evolutionary Computation); 2001
  3. 3. Kennedy J. Particle swarm optimization. In: Encyclopedia of Machine Learning. US: Springer; 2011. pp. 760-766
  4. 4. Jiang Y-L, Chen H-B. Time domain model order reduction of general orthogonal polynomials for linear input-output systems. IEEE Transactions on Automatic Control. 2012;57:330-343
  5. 5. Particle Swarm Optimization <http://mnemstudio.org/particle-swarm-introduction.htm>.
  6. 6. Salma U, Vaisakh K. Reduced order modeling of linear MIMO systems using soft computing techniques. Swarm, Evolutionary, and Memetic Computing. 2011. pp. 278-286
  7. 7. Zheng Y, Meng Y. The PSO-based adaptive window for people tracking. In: Computa-tional Intelligence in Security and Defense Applications. CISDA 2007. IEEE Symposium on, 2007; 2007. pp. 23-29
  8. 8. Basic PSO <http://tracer.uc3m.es/tws/pso/basics.html>
  9. 9. Venter G, Sobieszczanski-Sobieski J. Particle swarm optimization. AIAA Journal. 2003;41:1583-1589
  10. 10. Cervantes A, Galvan I, Isasi P. A comparison between the Pittsburgh and Michigan approaches for the binary PSO algorithm. In: Evolutionary Computation. The 2005 IEEE Congress on, 2005; 2005. pp. 290-297
  11. 11. Yao X, Liu Y, Lin G. Evolutionary programming made faster. IEEE Transactions on Evolutionary Computation. 1999;3:82-102
  12. 12. Xu F, Chen Y. A method for multiple sequence alignment based on particle swarm optimization. Emerging Intelligent Computing Technology and Applications. With Aspects of Artificial Intelligence; 2009. pp. 965-973
  13. 13. Shen Q, Shi W-M, Kong W. Hybrid particle swarm optimization and tabu search approach for selecting genes for tumor classification using gene expression data. Computational Biology and Chemistry. 2008;32:53-60
  14. 14. Choudhuri S. Bioinformatics for Beginners: Genes, Genomes, Molecular Evolution, Databases and Analytical Tools. US: Academic Press, Elsevier; 2014
  15. 15. Chuang L-Y, Tsai J-H, Yang C-H. Binary particle swarm optimization for operon prediction. Nucleic Acids Research. 2010;38:e128-e128
  16. 16. Molla M, Waddell M, Page D, Shavlik J. Using machine learning to design and interpret gene-expression microarrays. AI Magazine. 2004;25:23
  17. 17. Das S, Abraham A, Konar A. Particle swarm optimization and differential evolution algorithms: Technical analysis, applications and hybridization perspectives. Advances of Computational Intelligence in Industrial Systems. Berlin: Springer. 2008;116:1-38
  18. 18. Dirks RM. Analysis, Design, and Construction of Nucleic Acid Devices. California: California Institute of Technology; 2005
  19. 19. Poli R. Analysis of the publications on the applications of particle swarm optimisation. Journal of Artificial Evolution and Applications. 2008;2008:10. Article ID: 685175. DOI:10.1155/2008/685175
  20. 20. Selvan SE, Xavier CC, Karssemeijer N, Sequeira J, Cherian RA, Dhala BY. Parameter estimation in stochastic mammogram model by heuristic optimization techniques. IEEE Transactions on Information Technology in Biomedicine. 2006;10:685-695
  21. 21. Yeh W-C, Chang W-W, Chung YY. A new hybrid approach for mining breast cancer pattern using discrete particle swarm optimization and statistical method. Expert Systems with Applications. 2009;36:8204-8211
  22. 22. Wachowiak MP, Smolíková R, Zheng Y, Zurada JM, Elmaghraby AS. An approach to multimodal biomedical image registration utilizing particle swarm optimization. IEEE Transactions on Evolutionary Computation. 2004;8:289-301
  23. 23. Kaushik AC, Sahi S. HOGPred: Artificial neural network-based model for orphan GPCRs. Neural Computing and Applications. 2016;29(4):985-992
  24. 24. Kaushik AC, Dhar A, Sahi S. DrovePred: Server for DNA stem and BIME’s prediction using Particle Swarm Optimization. In: Bioinformatics and Systems Biology (BSB), International Conference on; 2016. pp. 1-5
  25. 25. Kaushik AC, Sahi S. Biological complexity: Ant colony meta-heuristic optimization algorithm for protein folding. Neural Computing and Applications. 2017;28:3385-3391

Written By

Aman Chandra Kaushik, Shiv Bharadwaj, Ajay Kumar, Avinash Dhar and Dongqing Wei

Submitted: 25 November 2017 Reviewed: 09 January 2018 Published: 13 February 2018