VarSearch: Annotating Variations using an e-Genomics Framework

José Fabián Reyes Román

1,2

, David Roldán Martínez

, Alberto García Simón

, Urko Rueda

and Óscar Pastor

Research Center on Software Production Methods (PROS), Universitat Politècnica de València, Valencia, Spain

Department of Engineering Sciences, Universidad Central del Este (UCE), San Pedro de Macorís, Dominican Republic

Department of Information Systems and Computation (DSIC), Research Center on Software Production Methods (PROS),

Spain, Valencia

Keywords: e-Genomics, EGF, GeIS, Variation, Conceptual Modeling, CMHG, Precision Medicine.

Abstract: Nowadays experts in the genomics field work with bioinformatics tools (software) to generate genomic

diagnoses, but the reality is that these solutions do not fully meet their needs. From the perspective of

Information Systems (IS), the real problems lie in the lack of an approach (i.e., Software Engineering

techniques) that can generate correct structures for data management. Due to the problems of dispersion,

heterogeneity and the inconsistency of the data, understanding the genomic domain is a huge challenge. To

demonstrate the advantages of Conceptual Modeling (CM) in complex domains -such as genomics- we

propose “VarSearch”, a web-based tool for genomic diagnosis that incorporates the Conceptual Model of

the Human Genome (CMHG) and takes advantage of Next-Generation Sequencing (NGS) for ensuring

genomic diagnostics that help to maximize the Precision Medicine (PM).

1 INTRODUCTION

The study and understanding of the human genome

could probably be considered one of the great

challenges of our century. Thanks to the advances in

NGS (Next Generation Sequencing) (Mardis, 2008),

there has been considerable growth in the generation

of genomic and molecular information. In addition,

the interactions that are available with this genomic

knowledge have a direct impact on the medical

environment and Precision Medicine (PM) (Grosso,

2016).

The application of Conceptual Modeling (CM)

(Olivé, 2007) techniques to the genomic domain

now provides solutions and optimizes some of the

processes carried out by experts (i.e., in genetic

laboratories and hospitals), and helps to solve the

problems that arise in handling the large amounts of

information from different sequencing methods. The

use of advanced Information System (IS)

engineering approaches can be useful in this domain

due to the huge amount of biological information to

be captured, understood and effectively managed. A

considerable part of modern Bioinformatics is

devoted to the management of genomic data. The

existence of a large set of diverse data sources

containing large amounts of data in continuous

evolution makes it difficult to find convincing

solutions (Reyes Román et. al., 2016). When we

addressed this problem from the IS perspective, we

understood that precise CMs were required to

understand the relevant information in the domain

and to clearly fix and represent it to obtain an

effective data management strategy.

Research and genetic diagnoses are typical

examples of the work done by experts -biologists,

researchers or geneticists- every day. However,

some information is required to perform these tasks.

Where are these data? Currently, this information is

dispersed in genomic repositories including web

sites, databanks, public files, etcetera, which are

completely heterogeneous, redundant, and

inconsistent (containing partial information). In

addition, most of these just focus on storing specific

information to solve a specific problem (e.g. YHRD,

designed to store Y chromosome haplotypes from

global populations, https://yhrd.org/). Due to these

characteristics, we are able to estimate the difficulty

of experts in finding and manipulating certain

genomic information, making this goal almost

impossible to achieve. Another relevant factor in the

domain is the constant growth and updating of the

328

Reyes Román, J., Roldán Martínez, D., García Simón, A., Rueda, U. and Pastor, Ó.

VarSearch: Annotating Variations using an e-Genomics Framework.

DOI: 10.5220/0006781103280334

In Proceedings of the 13th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2018), pages 328-334

ISBN: 978-989-758-300-1

data (i.e., biological concepts). The use of standard

definitions of concepts is not mandatory, so that

sometimes the same term can have different

definitions, in which case the meaning of the

concept depends on the interpretation given to it by

the expert. After studying this situation, we decided

to develop a Genomic Information System (GeIS) in

the form of the VarSearch tool (Reyes Román et. al.,

2018) for data treatment and management. This

system contrasts a set of genomic variations with the

information contained in a database that follows the

Conceptual Model of the Human Genome (CMHG)

(Reyes Román et. al., 2016).

Applying GeIS to the bioinformatics domain is a

fundamental requirement, since it allows us to

structure the Human Genome Database (HGDB)

with “curated” and “validated” data. The initial

research on applying CM approaches to the human

genome was reported in the works of Paton

(Bornberg-Bauer and Paton, 2002) and Ram (Ram

and Wei, 2004). The main goal in Ram's work was

to show the advantages and benefits of using

conceptual modeling to compare and search for the

protein in 3D (see other related works in (Pastor et.

al., 2010)). Reyes et. al. (2016) describes a CMHG

which proposes a domain definition at the

conceptual level. From this CMHG, we generated a

GeIS to support VarSearch. The application of CM

helps us to better understand and manage the

knowledge of the human genome.

This paper is divided as follows: Section 2 describes

the design and implementation of the VarSearch tool

as a GeIS. Section 3 describes two case studies

carried out using VarSearch. Finally, Section 4

contains our conclusions and outlines future work.

2 DESIGN AND

IMPLEMENTATION OF A GeIS

A GeIS can be defined as a system that collects,

stores, manages and distributes information related

to the behavior of the human genome. As mentioned

above, the GeIS described here is based on the

CMHG. This section deals with the VarSearch

design stage.

2.1 Design Overview

As outlined at Figure 1, VarSearch is based upon the

E-Genomic Framework (EGF), described in depth in

different research papers such as (Roldán et. al.,

2014). For the implementation of the tool, a series

of steps were carried out to ensure its good

performance, as explained below:

a) Human Genome Database (HGDB): The

transformation of the model defined for the database

schema (logical model) was almost automatic, using

the Moskitt tool (Muñoz et. al., 2010). In this task,

we found two different levels of abstraction in the

model. The conceptual model represents the domain

from the point of view of scientific knowledge,

while the database schema focuses on the efficient

storage and retrieval of data. For this reason, the

details of the physical representation must be

considered to improve the final implementation. It is

important to emphasize the integration of the two

tables "Validation" and "Curator" in the DB schema.

These tables are not actually part of the knowledge

representation of the domain, but are necessary for

the development and implementation of the tool.

Figure 1: EGF Framework and VarSearch (Roldán et. al.,

2014).

To load the HGDB the SILE methodology

(Reyes and Pastor, 2016) was used, which was

developed to improve the loading processes and

guarantee the treatment of "curated data". SILE was

used to perform the “search” and “identification” of

variations associated with a specific disease (a task

validated by experts in the genetic domain). When

the identified and curated data have been obtained

the "selective loading” is performed (through the

loading module) in the HGDB. The data loaded are

then “exploited” by VarSearch. Some of the diseases

(of genetic origin) studied and loaded were Alcohol

Sensitivity, Neuroblastoma, etcetera.

b) Selection of the different data sources: For

the choice of data sources, we addressed the

requirements raised in this first phase of the project.

After conducting studies and analysis of various

genomic repositories, we selected the following

databases: NCBI, dbSNP, UMD and BIC.

NCBI (https://www.ncbi.nlm.nih.gov/) is a data

source with curated data on structural concepts of

VarSearch: Annotating Variations using an e-Genomics Framework

329

DNA sequencing. From this repository, we extracted

information related to chromosomes, genes,

transcripts, exons... and everything related to the

"Structural view" of the CMHG. dbSNP (Sherry et.

al., 2001), BIC (Szabo et. al., 2000) and UMD

(Béroud et. al., 2000) are databases of variations that

store curated information on genetic differences

between individuals. The main reason for using

dbSNP is because it not only focuses on variations

of a specific gene or region, but also contains

variations related to all chromosomes and updates

the information immediately. BIC and UMD were

selected because of the requirements of a research

group that was collaborating with us in a project

(Future Clinic Project) focused on "breast cancer".

This group helped us to test the performance of the

GeIS and its associated tool.

c) Genetic loading module: For the loading

process of the HGDB, a load module was designed

to store the data from the previously measured data

sources. This load module was developed using an

ETL strategy (Zhou et. al., 2011) with three different

levels: extraction, transformation, and load (see

Figure 2). Each level is completely independent of

the others, facilitating and clarifying the design of

the system and improving its flexibility and

scalability.

Figure 2: Load module.

As can be seen in Figure 2, all the necessary

information is extracted from the source databases in

the first layer (1). All this raw unstructured data goes

to the second layer (2) where several transformations

are made to format the data according to the structure

of the database schema. These transformed data are

sent to the third layer (3), which communicates

directly with the database (following the above-

mentioned SILE methodology in Task “a”).

2.2 VarSearch Tool

VarSearch is a web application that allows the

analysis of variations obtained from the DNA

sequenciation of biological samples and which is

stored in FASTA or VCF file formats (Claverie and

Notredame, 2011). Different users can access the

application in private spaces in the HGDB and each

user can address his own variations. The validation

of variations that they consider relevant can be

included. It also offers storage for the users’

variations in order to find similarities in the file

analysis process.

Figure 3: VarSearch application.

Another advantage is the inclusion of the

information obtained from the data sources, together

with the user validations in the database, which is an

improvement in performance related to the search

for variations. VarSearch is capable of finding

variations in the database from a provided file (see

Figure 3). The variations found are displayed to the

user, and any additional information that the file

lacks can be calculated and validated. Any variations

of the file that have not been found in the database

can also be stored. After inserting one or more

variations not found in a file -because they are

considered relevant to the user- and reanalysing this

file, these inserted variations will be found in the

database and displayed to the user.

The VarSearch functionality has been grouped

into three main packages: (a) User management: a

user can act as administrator and control other users,

or can create new users and modify or eliminate

their Information. (b) Data load management: the

system allows the user to load the files to be

analysed in both VCF and FASTA format (Agliata

et. al., 2014), compare the variations in these files to

the variations in the HGDB used by VarSearch. (c)

Data analysis: After analysing and verifying the

variations in the input files, the user can list the

variations and classify them by multiple criteria

ENASE 2018 - 13th International Conference on Evaluation of Novel Approaches to Software Engineering

330

(position, chromosome, existence in the database,

etc.). There is also a series of functionalities related

to the login and modification of account information

that has not been grouped in any functionality

package.

2.2.1 Confidentiality of the Information

As this information is a company’s primary

resource, VarSearch restricts access to it. When a

user validates a variation, he can choose a privacy

category: (a) Public content, if he is willing to share

the knowledge with other users, or (b) Private

content, allowing access only to the owner-user and

hidden from other users. All the variations can only

be accessed by the user who created them.

2.2.2 VarSearch Architecture

In order to make it accessible to all users, VarSearch

was designed as a web application with HTML5

technology in a language common to all current

browsers. The information is managed by the

MySQL database. The VarSearch architecture

consists of the following elements:

(a) A distributable database based on MySQL

(using software tools like: Navicat Enterprise

and MySQL Workbench). For the initial

validation of this database, we only loaded the

information related to chromosomes 13 and 22.

(b) A set of REST services (Haupt et. al., 2014)

developed in Java using Hibernate and Jersey,

which are deployed on a Tomcat server 7.

framework for general organization of the

interface and files, together with jQuery to

define advanced interface components and

invoke REST services.

(d) It also includes a "mini" REST service to

manage users and roles, which is based on the

same architecture and technologies as the

other REST services. The data layer is based

solely on MySQL (you can see the VarSearch

architecture represented in Figure 4).

The application entry point is a file with

variations detected by a sequencing machine in VCF

or FASTA format. With this input the database is

searched to detect any variations, additional

information on the diseases they may cause and the

associated bibliography.

VarSearch users follow this process when

working with the tool:

• A VCF file is uploaded from the web.

Figure 4: VarSearch Architecture.

• The file is then processed and parsed. The

entries are shown on an HTML table and the

variants of each VCF entry can be seen.

• The variations present in the input file can be

annotated against the database and the

annotated file is downloaded in *.xls, *.csv

and *.pdf format or its contents viewed in

another HTML table.

To parse the VCF file and annotate the variants,

VarSearch relies on snpEff and snpSift (Tolhuis and

Wesselink, 2015) tools, and so well tested libraries

are used instead of reinventing the wheel. This also

ensures VCF standard support, using ANN files for

variant annotation. If another type of information is

considered useful for annotation and not covered by

the procedure described, VarSearch uses the “INFO”

field to introduce the desired values. As VarSearch

is based on EGF, new genome annotation files can

be quickly integrated by developing the proper

parser module, either by a custom development or

integrating a third-party tool or library.

All the information associated with the variations

found in the HGDB can be obtained. For variations

in the lists, user validations can be integrated for

future searches with the "Add Validation" option.

Another advantage of VarSearch is user

management (new users can be created and edited

using the "User Management" option). One of the

objectives of VarSearch is to continue the extension

and implementation of all the knowledge defined in

the CMHG, such as the treatment of pathways and

metabolic routes (Reyes Román et. al., 2017). This

tool facilitates the analysis and search for variations,

improving the generation of genomic diagnoses

associated with diseases of genetic origin. End-users

will find the web application easy to use and they

are guaranteed security for their data.

VarSearch: Annotating Variations using an e-Genomics Framework

331

Figure 5: Analysis of VCF file using VarSearch.

3 CASE STUDIES USING

VARSEARCH

To verify VarSearch performance, two case studies

were carried out. In the first, VarSearch was used to

analyse a VCF file. The second compared the time

spent on searching for variations manually and using

the application. To access the application VarSearch

users must have an account provided by Gembiosoft

SME (http://gembiosoft.com/). After logging in, a

file is selected for analysis. VarSearch reads all

records and transforms them into variations. These

transformations depend on the file information: for

example, the FASTA files contain a genetic

sequence (NG), and so require the reference on

which the variation is based to be to the “NG”

sequence. In contrast, VCF files use positions

relative to chromosomes (“NC”). Once the file

records have been converted into variations, the next

step is to search for these variations in the HGDB.

After the analysis, the “variations found” and

“variations not found” can be differentiated.

• Found Variations Management: Found

variations are those extracted from the file in

which information has been found in the

HGDB, which means that this variation has

been found in at least one genomic repository.

A found variation has much more information

than the variation obtained from the file and

allows us to calculate and submit detailed

information to the user. Having analysed the

VCF file, all the variations found are

displayed to the user, in each case calculating

the HGVS notation, its data source identifier,

clinical significance, and the number of

validations and databases found together with

their bibliographic references.

This information is calculated for VCF and

FASTA; however, VCF variations are sorted

by samples. Figure 5 shows the results

obtained by analysing a VCF file with a single

sample. For this sample (5323-BRCAyA), a

ENASE 2018 - 13th International Conference on Evaluation of Novel Approaches to Software Engineering

332

number of variations were found with the

corresponding information. A variation can

have validations made by users.

The validation column corresponds to the

number of validations that each variation has

and if a validation is private, only the owner

will see it. Another VarSearch feature is its

support for multiple bibliographical

references. A variation can be found in

different DBs and may contain different

bibliographic references.

• Not found Variations (insertion and treatment):

The user who is analysing variations may find

a variation in the file, which was not found in

the database. Using his experience and

knowledge he may consider some variations

as relevant despite not being found.

With VarSearch the user can insert the not

found variations or any variation considered

key to the study. If the user has inserted

certain variations that had not been found, on

reanalysing the file these inserted variations

are compared with the variations in the file,

showing the similarities. To differentiate the

variations of the different repositories from

user variations, the results obtained from the

user’s experience and the results from years of

study of different biomedical databases are

differentiated.

3.1 Improved Efficiency and Time in

Finding Variations with VarSearch

To validate the effectiveness and performance of the

proposed software, some experiments were

performed to measure efficiency and time. A study

was conducted to compare the time spent searching

for variations manually with an automatic search of

all the repositories mentioned above using

VarSearch.

Figure 6: Time optimization.

A manual search of one variation involves

detecting the variation in the VCF or FASTA file, a

search for the variation in the different databases,

and the identification and verification of the

variation. VarSearch was tested for the time it

needed to search for several variations, calculating

the time evolution according to the number of

variations involved (2, 5 and 7). The results can be

seen in Figure 6.

As can be seen in Figure 7, the cost of

performing a manual search rises to 5’32 minutes for

2 variations, 10’83 minutes for 5 variations and

18’89 minutes for 7 variations. However, with

VarSearch the time remains constant at between 2

and 3 seconds for different variations, which

confirms its efficient performance. Using this tool

thus significantly reduces the time spent on the

search for variations. Also, it must be remembered

that the manual search process does not calculate

additional information for variations. If this

information were necessary, the search time would

increase significantly, however, with VarSearch this

time remains constant because this information has

already been calculated in the search for variations.

4 CONCLUSIONS

VarSearch is a flexible new analysis framework or

web application that provides a powerful resource

for exploring both “coding” and “non-coding”

genetic variations. To do this, VarSearch integrates

VCF format input/output with an expanding set of

genome information. VarSearch (and other tools

built on EGF) will therefore facilitate research into

the genetic basis of human diseases.

EGF can also be expected to allow the

development of new tools in diverse e-genomics

contexts. As genetic laboratories are now oriented to

facilitating genetic procedures, web access, usability

and feasibility, the definition of different profiles are

therefore important goals. All this allows the user to

configure the tool according to his specific needs.

These necessities include inserting genetic variations

and validating its own variations, thus increasing its

“know-how”. VarSearch was tested in two different

case studies; one focused on the analysis of

variations (insert and search) and another to test its

search performance.

Future work will be oriented to the

implementation of -haplotypes and statistical

factors- (i.e., frequencies and populations) and

improving the next version of VarSearch (prototype)

for genetic diagnosis. Future research work will also

VarSearch: Annotating Variations using an e-Genomics Framework

333

be aimed at the application of Data Quality (DQ)

metrics to enhance the HGDB. We also intend to

extend the model with studies on the treatment of

“haplogroups”, including subjects with a similar

genetic profile who share a common ancestor.

ACKNOWLEDGEMENTS

This work was supported by the Generalitat

Valenciana through project IDEO

(PROMETEOII/2014/039) and the Spanish Ministry

of Science and Innovation through Project DataME

(ref: TIN2016-80811-P). The authors are grateful to

Jorge Guerola, Francisco Valverde, Ana León

Palacio, Ainoha Martín, Verónica Burriel Coll,

Mercedes Fernández A., Carlos Iñiguez-Jarrín,

Lenin Javier Serrano and Ma. José Villanueva for

their valuable assistance.

REFERENCES

Mardis, E. R., 2008. Next-generation DNA sequencing

methods. In Annu. Rev. Genomics Hum. Genet., vol. 9,

pp. 387-402, doi: 10.1146/annurev.genom.9.081307.

164359.

Grosso, L. A., 2016. Precision medicine and cardio-

vascular diseases. In Rev Colomb Cardiol, vol. 23, no.

2, pp. 73-76, doi: http://dx.doi.org/10.1016/j.rccar.

2016.01.026.

Olivé, A., 2007. Conceptual modeling of information

systems. Springer-Verlag Berlin Heidelberg, pp. 1-

445, doi: 10.1007/978-3-540-39390-0.

Reyes Román, J. F. et. al., 2018. Genomic Tools*: Web-

applications based on Conceptual Models for the

Genomic Diagnosis. In selected papers from ENASE

2017 in Communications in Computer and

Information Science (CCIS). Springer, pp. 1-21.

Reyes Román, J. F. et. al., 2016. Applying Conceptual

Modeling to Better Understand the Human Genome.

In Comyn-Wattiau I., Tanaka K., Song IY., Yamamoto

S., Saeki M. (eds) Conceptual Modeling. ER 2016.

Springer International Publishing, pp. 404-412, doi:

10.1007/978-3-319-46397-1_31

Bornberg-Bauer, E. and Paton, N. W., 2002. Conceptual

data modelling for bioinformatics. In Briefings in

Bioinformatics, vol. 3, no. 2, pp. 166-180, doi:

10.1093/bib/3.2.166.

Ram, S. and Wei, W., 2004. Modeling the semantics of 3D

protein structures. In Conceptual Modeling–ER 2004,

Proceedings. pp. 696-708, doi: 10.1007/978-3-540-

30464-7_52.

Pastor, M. A. et. al., 2010. Conceptual Modeling of

Human Genome Mutations: A Dichotomy Between

What we Have and What we Should Have. In

BIOSTEC Bioinformatics 2010, pp. 160-166, ISBN:

978-989-674-019-1.

Roldán M., D. et. al., 2014. An integration architecture

framework for e-genomics services. In 2014 IEEE

Eighth International Conference on Research

Challenges in Information Science (RCIS), doi:

10.1109/RCIS.2014.6861063.

Muñoz, J. et. al., 2010. Configuring ATL transformations

in MOSKitt. In Proceedings of the 2nd. International

Workshop on Model Transformation with ATL (MtATL

2010). CEUR Workshop Proceedings.

Reyes Román, J.F. and Pastor, O., 2016. Use of GeIS for

Early Diagnosis of Alcohol Sensitivity. In

Proceedings of the 9th International Joint Conference

on Biomedical Engineering Systems and Technologies,

vol. 3, pp. 284-289, doi: 10.5220/0005822902840289.

Sherry, S. T. et. al., 2001. dbSNP: the NCBI database of

genetic variation. In Nucleic acids research, vol. 29,

no. 1, pp. 308-311.

Szabo, C. et. al., 2000. The breast cancer information core:

database design, structure, and scope. In Human

mutation, vol. 16, no. 2, pp. 123, doi: 10.1002/1098-

1004(200008)16:2<123::AID-HUMU4>3.0.CO;2-Y.

Béroud, C. et. al., 2000. UMD (Universal mutation

database): a generic software to build and analyze

locus-specific databases. In Human mutation, vol. 15,

no. 1, pp. 86, doi: 10.1002/(SICI)1098-1004(200001)

15:1<86::AID-HUMU16>3.0.CO;2-4.

Zhou, H. et. al., 2011. An ETL strategy for real-time data

warehouse. In Practical applications of intelligent

systems, Springer Berlin Heidelberg. pp. 329-336, doi:

https://doi.org/10.1007/978-3-642-25658-5_41.

Claverie, J. M. and Notredame, C., 2011. Bioinformatics

for dummies. In John Wiley & Sons, pp. 1-456, ISBN:

978-0-470-08985-9.

Agliata A. et. al., 2014. IGV-plus: A Java Software for the

Analysis and Visualization of Next-Generation

Sequencing Data. In Vogiatzis C., Walteros J.,

Pardalos P. (eds) Dynamics of Information Systems.

Springer Proceedings in Mathematics & Statistics, vol

105. Springer, Cham, pp 149-160, doi: https://doi.org/

10.1007/978-3-319-10046-3_8

Haupt, F. et. al., 2014. A model-driven approach for REST

compliant services. In IEEE International Conference

on Web Services (ICWS), pp. 129-136, doi: 10.1109/

ICWS.2014.30.

Tolhuis, B. and Wesselink, J. J., 2015. NA12878 Platinum

Genome GENALICE MAP analysis report.

Reyes Román, J. F. et. al., 2017. Software Engineering

and Genomics: The Two Sides of the Same Coin?. In

Proceedings of the 12th International Conference on

Evaluation of Novel Approaches to Software

Engineering (ENASE 2017), pp. 301-307, doi:

10.5220/0006368203010307.

ENASE 2018 - 13th International Conference on Evaluation of Novel Approaches to Software Engineering

334