An FPGA-based Accelerated Mutation Detection System for the Tumor Suppressor Gene

This paper proposes a novel fast mutation detection system that looks for mutations in the tumor suppressor gene, also known as TP53. Mutations are modifications in the nucleotide sequences of the human genome and may be caused by various factors, such as exposure to radiation, sunlight, smoking and replication errors. Mutations in TP53 are the most common cause of cancer and early detection may prevent cancer from happening. The proposed system utilizes the high matching speed of a logic-based content-addressable memory (CAM) for mutation detection along with a hamming distance calculator that specifies the exact location of the mutation. The proposed system implementation is carried on a Xilinx Virtex®-7 field-programmable gate array (FPGA) and demonstrates a low match time of 0.18 μs, which is much faster compared to the state-of-the-art systems.


I. INTRODUCTION
T HE building block in every human's body is known as cells that perform different tasks. A combination of cells make the human body organs [1]. Each cell contains deoxyribonucleic acid (DNA); mostly located in the nucleus of the cells or rarely exists in mitochondria. DNA encodes the information required to build and maintain an organism (known as an organism's genetic blueprint) [2]. Entire human DNA comprises of around 3 billion bases [3], and above 99% of those bases are same in all the human beings. DNA composes of four chemical bases namely adenine, cytosine, guanine, and thymine or otherwise known as A,C,G, and T. Two strands pair up together with the help of a bond; the base pairs are A with T and C with G [4].
Nucleotide refers to the combination of each DNA base with a sugar molecule and phosphate. A spiral ladder-type shape is formed from a nucleotide which is known as double helix [5]. Genes are the small sections taken from the long chain of DNA. They are the essential unit of heredity (both functionally and physically) and in people, the number of DNA bases varies in each gene from a few hundred to over two million. Humans have approximately 20,000 to 25,000 genes, according to The Human Genome Project's approximation [6]. A mix of various genes together create chromosomes. Children acquire two sets of chromosomes; one from each parent. That is the reason why they have two duplicates of every gene. There are 23 pairs of chromosomes in human beings [7].
Genes contains the information regarding the formation of the molecules called proteins. It plays a vital role in cell functionality. Each cell depends on these proteins; present in thousands of number in order to perform the right task at the right time [8].
Mutated versions of genes known as alleles are created when a small-scale variation occurs in DNA bases' gene sequence. These variations in gene sequence can prevent the proteins from working properly [9]- [11]. Unable to perform the tasks can lead to change in a cell or an organ and leads to a disease. Thus, a variation of a gene can cause disease rather than the gene itself. Assume that if someone has "the cystic fibrosis" [12], the condition is caused by a variation of the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) gene. People without cystic fibrosis have a copy of the CFTR gene as well [13].
The most mutated gene in human cancer is T P 53. The T P 53 gene is known as tumor suppressor; it provides instructions for making tumor protein p53. It is located on chromosome 17 in humans and controls the cell division by keeping the cells from proliferating too fast. More than 50% of human cancers contain mutated T P 53 gene [14], therefore it plays a vital role in cancer prevention [15]. The basic point of the proposed work is to use content-addressable memory (CAM) for designing a fast mutation detection system, particularly for T P 53 gene. The proposed system displays high speed and throughput as compared to software-based systems [16], [17].
The key contributions of this research work include: • To the best of our knowledge, this is the first time a mutation detection system has been implemented using CAM on an FPGA device. • This mutation detection system performs better than the state-of-the-art systems in terms of power efficiency. • Along with very fast matching speed, our mutation detection system also specifies the exact location at which the mutation has occurred.
The rest of the paper is organized as: background study and related work are discussed in section II and III. Section IV focuses on computer memory and it's types. Using logicbased or slice register (SR)-based CAM architecture for mutation detection in human genome is discussed in detail in section V. The proposed system architecture and FPGA implementation are discussed in section VI. While section VII concludes this paper, having results, comparison and future work.

II. MUTATION IN HUMAN GENOME
The mutation is a change happening in DNA sequence, either by ecological factors like daylight, radiations, smoking, and so forth, or unintentionally during DNA replication [18]. These transformations in the human genome are otherwise called variations [19]. The human body comprises of 23 sets of chromosomes comprised of a long chain of DNA, which is practically equivalent to 3.2 × 10 9 base sets or 20-25k genes [20]. There are two subcategories in which each set of 23 chromosomes can be described; autosomes and sex chromosome, which are 1-22 and 23 rd respectively [21]. Various kinds of change can happen in these strands, going from revamping of the chromosome or duplication, or even a single nucleotide polymorphism (SNP) to the deletion of nucleotide [21]. Genetic diseases can be divided into two principle classes, i.e, chromosomal and mendelian (inherited) disorder.

A. CHROMOSOMAL DISORDERS
Typical vegetal or somatic cells go through a division cycle known as mitosis to accelerate their development, whereas meiosis division occurs in germ cells. Process reduction divi-sion (Meiosis I) and subsequently equational division (Meiosis II) are carried out during the meiosis. The development of a primary egg cell (or ovum) into a mature ovum, known as oogenesis, begins in the female embryo in the twelfth week, but it ends around the twentieth week at the phase of meiosis I, once the chromosomes have combined to form bivalents, quadruplicates, & replicate [22].
Misprints or inconsistencies in chromosome pairing, structural damage, and crossing-over are all causes of chromosomal disorders. Aneuploidy occurs when a cell has an abnormal number of chromosomes. A human cell, for example, may have 45 or 47 chromosomes instead of the usual 46. It is caused by abnormalities in the division of chromosomes during meiosis I and II [23]. Duplication, deletion, reciprocal T, translocation, Robertsonian T, and so on are included in chromosomal variations in human disorders. Gene-based mutation human disorders can be categorized as duplication, point mutation, deletion, splice mutation, insertion mutation, and dynamic mutation [21].
On the basis of functionality, the mutations can be subcategorized as, Gain of function mutation & Loss of function mutation. An increase in gene products or activity is caused by the gain of function mutation, which, in various cases, results in the creation of toxic products which lead to diseases like cancer, etc. On the other hand, the loss of function mutation results in the loss or decrease of gene products or activity.

B. MENDELIAN DISORDER
In light of some standards founded by Gregor Mendel in the nineteenth century, the patterns of inheritance characteristics or heredity are determined [24]. These principles include; autosomal dominant (AD), autosomal recessive (AR), X-linked recessive (XR), X-linked dominant and Y-linked disorders [21].
Moreover, few subjects having a disease that causes mutations don't show any signs and indications of the disorder, while others do. For instance, a mutation in the Breast Cancer type1 (BRCA1) gene may cause breast cancer in certain individuals while it probably won't be formed into the tumor in others [25].
Although certain genetic disorders do not follow any of the above-mentioned inheritance patterns. Mitochondrial illnesses including trinucleotide expansion abnormalities and genomic imprinting problems [26] are examples of non-Mendelian patterns of inheritance.

III. PRIOR WORK ON MUTATION DETECTION
This section of the study summarizes the most recent prior research in the field. As discussed in section II, change occurring in DNA sequence can cause several genetic disorders. Noval detection techniques are required to identify these mutations. Mutation detection has a vital role in genetic diagnosis including confirmational diagnosis, presymptomatic testing, and identity/forensic testing, etc.
In order to discuss the recent work done in the field of mutation detection, prior work can be categorized into two sub-categories, i.e, software-based and hardware-based systems. In software-based algorithms, [16] presented three algorithms: First-Last Pattern Matching (FLPM) algorithm, Processor-Aware Pattern Matching (PAPM) algorithm and Least Frequency Pattern Matching (LFPM) algorithm. FLPM algorithm performs character-based comparison, in which the first and last indexed character of the query string is compared with the stored pattern. The query string window comprises of six characters. In order to reduce the window size and to improve run-time, word-based comparison (a word comprised of few characters) approach is used in PAPM and LFPM algorithm.
LFPM [16] surpassed the other two algorithms, in term of time required for matching and reducing query pattern window size. As LFPM only look for the word of query string in the stored string that has the lowest occurrence frequency. The main disadvantage of these three algorithms is that it don't perform parallel operation. Secondly, the shifting of query pattern window over a particular span of stored pattern also requires more memory.
Discrete to Continuous (DTC) algorithm is developed and presented in [17], in which superposition of a discrete test points over continuous reference points is performed. This algorithm is also based on character matching approach. The pre-processing of searching the superposition of test pattern in reference pattern is very time consuming. The nucleotides of test pattern and reference pattern is represented by an abscissa and an ordinate in a space of 2 dimensions (2D). The issue with this algorithm is that the distance calculation between the test pattern and reference pattern is done by nonlinear equation (having square-root), which makes its hardware implementation nearly impossible because it exploits the hardware resources.
The hardware-based system (for mutation detection) presented in [27] is using approximate string matching (ASM) approach. The hardware is named as genome approximate string matching (GenASM). Main disadvantage of GenASM is that it is a random access memory (RAM) based system, which performs sequential computation and thus it lacks in parallel search capability.
Therefore, a question arises, is there any way to tackle the issues in the aforementioned systems? The answer is yes. The proposed system uses CAM which has the capability to compare the test pattern concurrently with the reference; thus reducing the execution time. Furthermore, the system presented in this paper is also resource efficient. Architecture and methodology of the proposed system are explained in detail in section IV, V and VI.

IV. COMPUTER MEMORY & CLASSIFICATION
Computer memories are designed to consume less power or read/write data at a high speed, depending on discrete requirements or parameters. Additionally, on the basis of data storage i.e. permanent or temporarily depending on the power, computer memories can be categorized as volatile and non-volatile [28]- [30].
These days, several types of memories exist however primary memory and secondary memory are the basic type of memories. System memory is actually the primary memory while storage is referred to as secondary memory. Memory is further categorized as CAM, read-only memory (ROM), and RAM. As the proposed system uses CAM; therefore section IV-A focuses on CAM, its types, and how it differs from RAM.

A. CAM, ITS TYPES, & COMPARISON WITH RAM
In contrast to RAM, which takes an address as an input and outputs stored data, a CAM's input takes data and the output gives the stored data addresses as shown in Fig. 1. Due to parallel processing being carried out in CAM, its speed is faster than RAM. On the basis of functioning, CAM is divided into two types as stated below;

1) Classical Content Addressable Memory (CAM): In
CAM, every CAM cell typically contains a storage circuit and a comparator. When the search data is passed to the CAM cell, the entire memory is simultaneously searched, and the output with the highest priority is selected using an encoder. Fig. 2 illustrates the two types of traditional CAM, namely (a) Ternary Content Addressable Memory (TCAM) and (b) Binary Content Addressable Memory (BiCAM), based on the operation mechanism. For exact matching, "BiCAM" is used since it can only store "0" or "1" but "TCAM" is used for partial matching because it can store a third state, which is symbolised by the letter "X" and is known as the "don't-care bit". The mask bit in TCAM is represented by the 'X' bit; an exact match will occur if the mask bit value is '0', similar to a BiCAM, but if the mask bit value is '1', the match will always appear as 'true', regardless of the search query supplied. CAM is now used in a variety of applications, including pattern matching, data compression, packet categorization, and image processing [31]- [34]. 2) Field Programmable Gate Array (FPGA) based CAM: In the 1980s, a programmable arrangement recognized as Programmable gate array (PGA) or Programmable logic array (PLA) later on called Programmable logic devices (PLDs) were introduced with some programming and employing AND-OR structures. FPGA, VOLUME 4, 2016 which was initially conceived by Xilinx, is created by combining programmable ROM (PROM) with PLD [35], [36]. The FPGA is built in such a way that it may be reprogrammed by the designer after it has been manufactured. FPGAs are frequently configured using the hardware description language (HDL). It's used for developing and testing digital circuits [37]. Because of fact that CAM is an expensive memory as compared to the RAM, therefore it's functionality can be emulated using FPGA hardware. Remember, there is no build-in CAM input port on FPGA though we can use logic gates available on the FPGA to emulate CAM functionality. Static RAM (SRAM)-based CAM, logic (flip-flops)-based CAM, slice register (SR)-based CAM and Look Up Table (LUT)-based CAM are the types of FPGA based CAM.

V. METHODOLOGY
The requirement of large memory units for huge biological data and high processing time are the issues faced by Next Generation Sequencing (NGS) techniques [38], [39]. In order to address the processing time issue, CAM is used for the detection of mutations in human genome and a single nucleotide polymorphism (SNP) in DNA sequence is targeted in this research work. Accuracy, expensive fast computation and memory requirement, are the issues faced by already developed SNPs detection algorithms [40]. Therefore, a fast, accurate and scalable mutation detection system is proposed.
The proposed system is much speedy than RAM based system due to the concurrent search/match mechanism of CAM.
Where the exact matching is achieved by using SR-based or logic-based CAM architecture [32]. The T P 53 gene sequence is taken into consideration in the proposed system to detect mutations, as it mutates in most types of human cancers [41]- [44]. T P 53 gene codes or provides instructions for making a tumor protein known as p53. It functions as tumor suppressor protein and controls the proliferating or an uncontrolled cell division process. p53 is altered in most cancers; therefore, a well-timed diagnosis of the mutated T P 53 gene is essential.
Moreover, to recognize the number of nucleotides mutated in the particular section of the genome, a hamming distance calculating algorithm is used. Where the test pattern is compared with the benchmark datasets. Point mutation is a type of mutation where a single nucleotide is embedded, changed, or erased from the DNA chain. This extra feature of hamming distance calculator actually helps in detecting the point mutation.

A. THE GUARDIAN OF THE GENOME, P 53
T P 53 is the gene that codes for the tumor suppressor protein "p53". Every cell in the human body contains this tumor suppressor protein. It guards against any kind of damage to DNA. Poisonous compounds, smoking, destructive radiation, and bright ultraviolet (UV) beams are just some of the things that can harm DNA. The p53 continuously checks for the damaged DNA as it is directly tied to the DNA chain. If any damaged part is found in the DNA chain, p53 assumes a fundamental part in fixing it. It either informs the respective cell to undergo self-destruction/eliminating the damaged part or try to repair it. The p53 activates other genes to fix the damaged part of the DNA if it is repairable else it stops the cells from cell division (mitosis). Accordingly, it forestalls the development of tumors.
The tumor protein (p53) is also known as "the guardian of the genome", as it monitors the DNA chains, controls the mitosis process and can repair the genome if required. The term p53 refers to the role it plays in preventing changes to the human genome. Fig. 3 depicts the entire utility of p53. The p53 gene, as shown in the Fig. 3, deals with the issues depicted by the arrows pointing toward it, as well as the strategies depicted by the arrows pointing away from it. All these functionalities of the p53 contribute towards tumor suppression. Freely accessible benchmark datasets of the national center for biotechnology information (NCBI) and the international agency for research on cancer (IARC), have been used to acquire the T P 53 gene or p53 protein succession [45]. The combination of three nucleotides results in the formation of a codon. There are 394 codons in the p53 genomic sequence. In the revised reference sequence, codon position 72 CCC (pro) is mentioned, whereas CGC (arg) is indicated here [45]. This data (genome sequence) is then mapped into an SR-based CAM.

B. ARCHITECTURE OF SR-BASED CAM
In this research work, we employed logic-based or SR-based BiCAM to achieve quick and precise genomic pattern matching. That simulates the capability of a logic-based higher performance binary CAM (LH-CAM) [32], but instead of sending matched addresses to the output [32], the address of the mismatch will be provided by the CAM in our system. The SR-based CAM used in our system is shown in Fig. 4.

C. THE P53 GENOMIC SEQUENCE MAPPING
Adenine (A), cytosine (C), guanine (G), and thymine (T) are the four nucleobases (also known as nitrogenous bases) that make up DNA. As shown in Fig. 5, to create double-stranded DNA, all four bases are coupled together (for example, C with G and A with T). Since DNA has four bases in total, we could use Eq 1 to encode individual bases with two binary bits.

Total number of bases
"b" denotes the total number of nucleotides, and "m" denotes the number of bits per nucleotide. If each base is represented by two bits, m = 2, then b = 2 2 = 4. Two flip-flops (FFs) or MUs are required in the SR-based CAM to map a single base or nucleotide. The binary representation of each base is shown in Table 1.

Nitrogenous Bases
Adenine Cytosine Guanine Thymine (A) (C) (G) (T) Representation in Binary 00 01 10 11 A codon is a sequence of three nucleotides. The "p53" sequence is made up of 394 codons [45], totaling 1182 nucleotides. Since each nucleotide is represented by two binary bits, the whole sequence of "p53" requires a total of 2364 bits to map.
Accordingly, the CAM is 66x36, with 66 horizontal rows (MLs) and each row consisting of 36 bits or 36 MUs. Hence, it has 66 vectors (Vs), each of which is 36 bits long. The mapping of the T P 53 gene sequence to the SR-based CAM is shown in Algorithm 1. The codons are distributed as 6 codons per ML to maximize resource use. Every horizontal ML has 18 nucleotides, requiring 36 bits per ML to map them. Table 2 shows a mapping example using Algorithm 1. The mapping difficulty is O(1), although in prior CAM architectures, the updating difficulty increased exponentially with the depth of the CAM and also required pre-processing [46]. Consequently, SR-based CAM is easy and quick to update.

D. THE MATCHING MECHANISM
Each MU has the ability to store and compare a single bit. The test pattern or query string (QS) is compared concurrently with all the vectors (Vs) using a bit-by-bit comparison procedure. The matching/searching difficulty is also "O(1)" due to concurrent processing.
. Map n Codons Ensemble to Vi of the SR-based CAM 6: end for As every MU has a single bit comparator that compares the QS to the stored Vs bit by bit, if a single bit mismatch happens in any ML, the entire ML turns low (logic 0), however if the bits stored in Vs match the input QS, the entire ML remains high (logic 1). The mismatch-line address (MMA) is delivered to the output if there is a mismatch; else, default address (DA) is delivered to the output. 6 codons or 18 nucleotides are available in the first match-line (M L 0 ), for example. Table 2 shows its binary representation, and Fig. 6 illustrates the matching procedure. Due to the 2 nd nucleotide in the 3 rd codon, a mismatch occurred in M L 0 . As a result, the entire match-line becomes low (logical 0), and MMA is passed to the output (in this example the mismatch-line address is 0).
As illustrated in Fig. 7, the proposed system is made up of three primary blocks; SR-based CAM and matching/mismatching have already been explored extensively in the preceding paragraphs. The test sequence is fed into SR-based CAM, which employs Algorithm 2 to detect the mutation by comparing it to the benchmark stored sequence. As per the architecture of SR-based CAM, each MU has 1- Match occurs and DA is sent to output 6: else 7: Mismatch occurs and MMA is sent to output 8: Respective QS ⊕ Vi stored at MMA 9: Number of mutated nucleotide extracted 10: end if bit storage and a 1-bit comparator. Thus, the search vector is compared bit-by-bit with the stored vector at every ML. If a single bit mismatch occurs, the entire ML will turn to a low state and the system will indicate that a mutation has been found. The mutated pattern is then given to the hamming distance calculator, which calculates the number of altered nucleotides. If there has been no mutation found, the system will return the default value, DA.

E. HAMMING DISTANCE CALCULATOR
The hamming distance calculator compares the reference benchmark pattern to the mutated or test pattern. As shown in Fig. 8, the number of locations where the QS differs from the SP is the Hamming distance between two vectors (test pattern (QS) and stored pattern (SP ), d(QS, SP ).
In order to elaborate the functionality of the hamming distance calculator, let suppose a mismatch occurs at a particular ML. The test vector at that ML will then be provided to the hamming distance calculator and bit-by-bit comparison (XOR) is perform with the reference vector. The priority encoder will receive the outputs of all the XORs in order to determine the exact number of altered nucleotides.

VI. ARCHITECTURE & FPGA IMPLEMENTATION
For implementation, Verilog HDL is employed. While Xilinx's ISE Create Suite 14.5 is utilized for design purpose. The proposed system is implemented and tested on the Virtex-7 device XC7VX330T FPGA using the Xilinx simulation tool. For the analysis, post-place and route results are taken into account.

A. ARCHITECTURE OF PROPOSED MUTATION DETECTION SYSTEM
The FPGA-based mutation detection system is built with a 66x36 size SR-based CAM and encoding circuits. The depth of the proposed mutation detection system is 66, with 36 bits stored in each match-line vector. Logic gates, a comparator, and encoding circuitry make up the overall system. The proposed system's I/Os have four input ports (CLK, in, QS, and set) and one output port (EN CRes), as shown in Fig. 9. The clock is represented by "CLK", the reference benchmark sequence is represented by "in", the query string is represented by "QS", and "set" acts as a switch to activate a certain match-line. While "EN CRes" (encoder result) is the output port that displays the address of the mismatch. In addition to all of this, the hamming distance is computed and shown on the console.

VII. RESULTS AND PERFORMANCE COMPARISON
For performance comparison, based on the most relevant research work. A comparison is made between the proposed system and recent research [16], [17], [27].

A. SPEED OF THE PROPOSED SYSTEM
The proposed system's CAM is made up of SRs, LUTs, and logic gates. Furthermore, it can complete the matching operation in just two clock cycles. Thus it improves speed; the proposed system performs better in terms of match time. The detection time of various software and hardware-based systems is compared in Table 3. GenASM [27] Hardware based 64bp -320bp 0.00225 ms 3.
DTC algorithm [17] Software based 1000 (nucleotides) 17 ms The proposed system has 66 match-lines, and the matching process takes 2.803 nanoseconds per match-line. Because of the usage of CAM, all of the processes are completed simultaneously, resulting in a system speed of 356.7 M Hz. Table 4 provides the details.

B. POWER CONSUMPTION
According to the XPower Analyzer results, the presented technique consumes 27 mW of dynamic power. The power consumption of the GenASM system [27] built on FPGA is compared with the presented system. GenASM is a first-ofits-kind ASM-based genomic sequence analysis system. The GenASM hardware is made up of on-chip SRAMs, whereas the proposed system hardware is entirely made up of SRs and LUTs. They employed SRAMs to simulate parallel computing capabilities; however, the proposed system uses SR-based CAM to provide the same capability while also being faster and more power-efficient. According to the results in Fig.  10, it is 3.74 times more power-efficient than the GenASM. The GenASM uses 101 mW of power and has been tested with DNA base pairs ranging from 64 to 320. On the other hand, the proposed system has been tested on 1,188 DNA base pairs.

C. RESOURCES UTILIZATION
An FPGA resources like Block RAMs, SRs, LUTs are used for logic implementation, and so on to construct the system. Table 5 lists the resources used for the proposed system, as well as the resources that are available. Table 5 shows that the proposed system is resource efficient, since it uses just 1% of the SRs and LUTs. After the system's post-place and route implementation, the resource consumption report is generated. For rapid and efficient VOLUME 4, 2016

VIII. CONCLUSION & FUTURE DIRECTIONS
Parallel processing, often known as parallel computation, is vital in the detection of mutations. The application of CAM to the detection of genomic mutations is a novel method. This approach will contribute to the development of a fast, power-efficient, accurate, and cost-effective mutation detection system for the entire genome in the future. In comparison to recently published work, the results indicate that the proposed system is power-efficient (consumes 27 mW ), has minimal execution time (0.18 µs), and is substantially faster (356.7 M Hz). This can contribute in the early detection of various diseases, including cancer and viruses that adapt and evolve as they infect and proliferate in their hosts' cells. In the current pandemic situation of COVID'19, this system can help in identifying the mutation of coronavirus. Moreover, the proposed system is FPGA-based, which means that it can be reconfigured. In the future, we may be able to construct a genome approximate matching method or system. Engineering Co. Ltd. His research interests include low power/high speed CAM design on FPGA, re-configurable computing, pattern recognition, embedded systems, and image processing. He has around 50 Journal and Conference papers and has got 6 patents (US and Korean) on his name in the field of FPGA-based TCAMs. He has supervised 2 PhD students and more than 20 MS students. He is a reviewer of many prestigious journals such as IEEE Transactions on Very Large-Scale Integration Systems, IEEE Transactions on Circuits and Systems I, IEEE ACCESS, Circuits, Systems, and Signal Processing, and IEEE Embedded Systems Letters. He was awarded scholarship for MS studies in Hanyang University, South Korea by Higher Education Commission Pakistan, and for PhD studies by City University of Hong Kong. He is an expert in Outcome-based Education system (OBE) of the Washington Accord. He also received two times best researcher award (CECOS University, Pakistan), two times Outstanding Academic Performance Award, two times Research Tuition Fee Scholarship, and two times Conference Grant (City University Hong Kong). Further, he has also completed 4 months training on "Fachhochschule Teaching and Management" in two Austrian universities -FH-Joanneum, Graz, and MCI, Innsbruck.
TAMA FOUZDER received Ph. D. in Electronic Engineering Department from City University of Hong Kong in 2016. She received her Bachelor's degree in Electronics and Communication Engineering Discipline from Khulna University of Bangladesh in 2006. She worked as a research assistant in the EPA centre, Electronic Engineering Department of City University of Hong Kong. She has around nine years teaching experience as a Lecturer in the Electronic and Telecommunication Engineering Department of University of Development Alternative, Bangladesh. At present, she is working as an Assistant Professor in Electrical and Electronic Engineering Department of University of Liberal Arts Bangladesh, Bangladesh. Her research interest includes electronic product reliability and electronic packaging, lead-free nano composite solders and final surface finish, FPGA based design, antenna design, and biomedical engineering. She has published ten (10) international journal papers and ten (10) peer-reviewed international conference papers. She served as a reviewer for some international journals and international conference papers. VOLUME 4, 2016