=Paper=
{{Paper
|id=None
|storemode=property
|title=Facing the Challenges of Genome Information Systems: a Variation Analysis Prototype
|pdfUrl=https://ceur-ws.org/Vol-592/PaperDemo02.pdf
|volume=Vol-592
|dblpUrl=https://dblp.org/rec/conf/caise/MartinezMVVLP10a
}}
==Facing the Challenges of Genome Information Systems: a Variation Analysis Prototype==
https://ceur-ws.org/Vol-592/PaperDemo02.pdf
Facing the Challenges of Genome Information
Systems: a Variation Analysis Prototype.
Ana M. Martinez, Ainoha Martín, Maria José Villanueva, Francisco Valverde,
Ana M. Levín, and Oscar Pastor
Centro de Investigación en Métodos de Producción de Software
Universidad Politécnica de Valencia
Camino de Vera S/N 46022, Valencia, Spain
{amartinez, amartin, mvillanueva, fvalverde, alevin, opastor}@pros.upv.es
Abstract. In Bioinformatics there is a lack of software tools that t with
the requirements demanded by biologists. For instance when a DNA sam-
ple is sequenced, a lot of work must be performed manually and several
tools are used. The application of Information Systems (IS) principles
into the development of bioinformatic tools, opens a new interesting re-
search path. One of the most promising approaches is the use of concep-
tual models in order to precisely de ne how genomic data is represented
into an IS. This work introduces how to build a Genome Information
System (GIS) using these principles. As a rst step to achieve this goal,
a conceptual model to formally describe genomic mutations is presented.
In addition, as a proof of concept of this approach, a variation analy-
sis prototype has been implemented using this conceptual model as a
development core.
1 Introduction
Thanks to the breakthrough of the Human Genome Project and the advances
in DNA sequencing, an enormous amount of genetic data is being produced by
researchers every day. Most of these experiments are focused on the understand-
ing of the relationship between genotype (gene con guration and combination
of a particular individual) and phenotype (expression of the genes in a speci c
human feature). As a consequence, the creation of biological databases and tools
to exploit the produced data have grown drastically. However, these tools and
databases have usually been de ned to support an speci c research area or ex-
periment. Therefore, when biologists want to use them for a particular assay, it is
very unlikely that they support their speci c requirements. This issue leads to a
situation where the researcher has to spend a lot of time and eort to perform a
simple analysis. Since these bioinformatics tools are not developed using IS prin-
ciples, they are not aligned with the user requirements. The main consequences
of this issue are:
Some biological databases are only human readable, thus cannot be processed
properly in an automatic way.
The extraction of relevant data is di cult because it is spread around dif-
ferent databases.
Since several tools are required to analyze the data, the speci cation of the
tooling work ow and integration is far from trivial.
Inclusion of new studies and bibliography into the available tools turns into
a hard task.
With the goal of facing these issues, some researchers have proposed [1] the
development of Genomic information Systems (GIS), an IS speci cally designed
to handle a big amount of genomic data. In this work, a new approach to de-
velop GIS is proposed: the use of conceptual models to organize genomic data
and guide the development. Thanks to the close collaboration with biologists in
the context of this research project, the gap between the disciplines of Software
Engineering and Genetics is solved. The result of this interdisciplinary collabo-
ration is a conceptual model that guides the alignment of concepts among both
elds. Therefore the design and implementation of the software artifacts that
made up a GIS becomes an easier process.
Following that idea, this paper presents a GIS prototype that analyzes a DNA
sequence in order to nd documented variations for a speci c gene. Once all
variations are located in the sequence, the prototype splits them in two groups:
one group contains harmless variations and the other one contains variations
that produce a change in gene or protein function. For those in the last group,
their speci c phenotype is reported as it has been described in the literature.
This information is bibliographically referenced and gathered in a report that
helps the researcher to understand the genetic meaning of the variation and why
it produces a certain phenotype. This is very useful because it can speed up
the diagnosis of a speci c disease. Furthermore, it is widely accepted that an
early disease detection might be determinant. The main contribution of this
work is that the GIS development is supported by a set of conceptual model
entities that formalize the domain concepts related with genomic variations.
As a consequence, the conceptual model plays an integration role to provide
the genomic knowledge in an unambiguous way and independent from speci c
datasource details.
With that goal in mind, the rest of the paper is organized as follows. In
section 2 a review of DNA variation analysis tools is presented. Section 3 details
a conceptual model for describing genomic variations. Section 4 describes how the
variation analysis prototype has been developed. Finally, in section 5 conclusions
and future work are stated.
2 Related Work
In recent years, several commercial tools have been developed to provide genomic
analysis. These tools can perform tests in order to estimate the customer proba-
bility to suer certain diseases. Navigenics [2], 23andMe [3] and deCODEme [4]
are the most relevant tools in this eld. The dierences between them are brie y
summarized in Table 1. However, the accuracy of these tools is far from ideal.
Results are not reported in an unambiguous way because biological concepts
are not precisely de ned. Without a conceptual model that guides the precise
de nition of the domain, further integration with external tools is complex to
achieve.
Another drawback of these tools is that the only variations reported are
SNPs (Single Nucleotide Polymorphism). The conceptual model improves the
reports quality because other complex variations such as repetitive insertions or
deletions are classi ed. Furthermore the diseases detected by these commercial
tools are constrained to the number of supported genes. The use of a conceptual
model overcomes this constraint because provides guidelines to support several
gene sequence references and their new discovered variations.
Table 1. Comparison of DNA analysis tools.
Navigenics 23andMe deCODEme
Analysis Type Genotyping Genotyping Genotyping
Platform Aymetrix [5] Illumina [6] Illumina [6]
Variations (million) 1 (only SNP) 0.5 (only SNP) 1.2 (only SNP)
Detected diseases 28 51 49
3 A Conceptual Model for Describing Variations
The main objective of the conceptual model presented in this paper is to es-
tablish a connection point between the genomic eld and the GIS development
domain. One of the main characteristics of the genomic eld is heterogeneity.
The uni cation of the relevant concepts is a di cult task, since genomic con-
cepts are not precisely de ned. Moreover, the eld knowledge is still developing
and these concepts are constantly evolving, making the organization of all the
genetic data available more di cult.
Genetic databases are thus aected by this heterogeneity problem. Each
database re ects the concepts according to the interpretation and terminology
of a biologist. However, there are dierent de nitions for the same concept; for
example, a variation in the DNA sequence is referred under the terms: variation,
mutation, polymorphism or SNP [7]. Even though all of them represent more or
less the same concept, there are slight dierences among them. The problem of
heterogeneous data can be solved with the use of conceptual models, as some
works have proposed [8]. The development of a conceptual model to represent
the human genome is a useful approach to understand this complex domain since
precise concepts are de ned and related among them. If new concepts, relations
or changes are discovered, they can be easily incorporated into the model.
The conceptual model presented here claims to be precise with genetic con-
cepts and IS principles because it has been developed by software engineers and
biologists specialized in the genomic eld. The model presented in this section
is focus on the description of genomic variations. However, it is an excerpt of
a widest one [9], whose main goal is the speci cation of the required human
genome concepts for developing GIS.
Fig. 1. Conceptual Model for describing variations
Figure 1 shows the proposed conceptual model. At the top of the picture (1)
Gene and Allele modeling entities are de ned. Gene entity models the generic
concept of gene whereas Allele entitiy represents the individual instances of a
gene. The Allele entity has two specializations: Allelic Reference Type and Allelic
Variant. Allelic Reference Type models the reference sequence that de nesa "uni-
versal" gene to be used for comparison purposes. These reference sequences are
extracted for trusted data sources as RefSeqGene database [10]. Allelic Variant
represents a DNA sequence of an individual which has several variations from
the allelic reference.
Each variation discovered by means of the comparison process performed over
a sequence, is modeled by the Variation entity (2). The Variation entitiy stores
all the variations documented in the genetic literature that are associated to some
disease or to normal changes because of the intrinsic nature of an individual. This
entity has two specializations: Precise variations, which de ne a variation that
is completely located and Imprecise variations, whose location details are not
speci ed. Precise variations are also categorized in four entities according to the
change performed in the sequence : Insertion, Deletion, Indel (insertion/deletion)
and Inversion. An indel can be categorized as SNP as well when it occurs at
least in 1% of the population.
A variation that is speci ed in the model is always related to its phenotype,
which is modeled by the Phenotype entity (3). The Certainty entity speci es
the probability that a phenotype could show up because of a concrete variation
on the genotype. In case is identi ed a genotype-phenotype association, it is
essential to know information about the bibliographic reference and the original
database where the discovery was stated. This data is de ned by the Bibliog-
raphy Reference and BibliographyDB entity (4) respectively. As a rst result of
this conceptual model, a genetic database (GDB) has been created to store the
variation information that is used by the presented GIS prototype.
4 A GIS Proof of Concept: a Variation Analysis
Prototype
The main goal of the prototype is to show how conceptual models can be useful
to de ne a GIS. One of the most common tasks in the genomic area is the anal-
ysis of the genomic sequences [11]. Researchers perform the analysis by doing a
comparison between a certain DNA sample from a concrete gene and its refer-
ence sequence. The comparison is done using an alignment tool that shows a list
of dierences among them. After that, an experienced researcher has to decide
which variations are relevant and which not. Then, they have to dive into the
vast and non-structured amount of information that is scattered across the Web
and search the bibliography that justi es each relevant variation. Performing
this work manually is a tedious and time consuming task.
The proposed prototype reduces this time by automating the major part of
the manual work. This automation can be done thanks to the conceptualization
of the domain by the presented conceptual model. Data such as genes, variations,
phenotypes and bibliographic references is now represented as perfectly de ned
conceptual entities. Thanks to this conceptualization, heterogeneity and data
dispersion problems are solved, avoiding the manual preprocess of some non-
computer legible data and ensuring the quality of the data stored.
Fig. 2. Prototype Phases
The purpose of the presented GIS prototype is to receive a DNA sample
from a patient and provide a report that helps the doctor to diagnose a certain
disease. The experts only have to introduce the sample in the suitable format
and review the provided results, forgetting everything about manual treatment
and endless searches across the bibliography.
The analysis process performed by the prototype is summarized in gure 2.
Some conceptual model entities that are used in the dierent steps are depicted
in white rectangular boxes. The process is divided into ve main steps:
1. Input data: The biologist selects a gene from the set supported by the pro-
totype, for instance the BRCA1 gene, and introduces the DNA sample to
be analyzed. The input of the sample can be performed manually or by
uploading a le in FASTA format.
2. Alignment report: According to the selected gene, the prototype locates the
suitable reference using the allelic reference entity. After that, an alignment
process between the sample and the reference is carried out for nding varia-
tions. This alignment is performed using the BLAST algorithm [12], however
importing results from DNA sequencing tools as Sequencher [13] will be sup-
ported in next versions. Using the de ned conceptual model, each discovered
dierence is formalized as an instance of the variation entity. This formal-
ization, which it is not present at the moment in other tools or databases, is
independent of the output from any alignment tool and provides a suitable
way for exchanging variations. A report that summarizes all the changes is
generated using these variation entities.
3. Variation knowledge: Thanks to the report generated in the previous phase
the classi cation problem is simpli ed. Variations are located according to
a well-know reference sequence and their positions match with the genomic
data stored in GBD. Then, each variation is queried into the GDB to deter-
mine if it has been de ned as a precise variation. If a variation cannot be
found in our GDB is classi ed as unknown. At this point, known variations
are classi ed into an speci c type of sequence change. Unknown variations
are classi ed as non-silent if the variation produces a change, in other words,
an eect in the expected gene product (protein).
4. Phenotype Assessment: Variations classi ed as known may have some phe-
notype associated. In order to asses if the phenotype is related to an speci c
disease, a research publication is required to provide a trustful evidence. For
those cases, the conceptual model describes the bibliographical reference that
supports the phenotype for an speci c variation. In the context of this work,
variations with a pathogenic phenotype are classi ed as mutations whereas
they are classi ed as SNPs if no negative phenotype is described.
5. Report creation: All the obtained information is gathered in a report. This
report contains information about the variations found: mutations, variations
whose phenotype is not a disease and unknown variations. Each variation is
provided with the following information: the location where it was found in
the sequence, its type (Insertion, Deletion, Indel or Inversion) and the num-
ber of nucleotides inserted or deleted. For the mutations found in the GDB
their associated phenotype and its bibliography is added as well. Finally, the
report le can be saved as a text document.
5 Conclusions and Future Work
This work proposes a GIS engineering solution in order to solve the problems of
heterogeneity on the genomic domain. A conceptual model is presented which
describes and de nes formally the concepts related to genomic variations. As a
proof of concept, a GIS prototype, which uses this conceptual model as back-
ground, has been implemented.
One of the advantages of using the presented GIS prototype is that the vari-
ation analysis can be performed using only one tool, avoiding the data work ow.
In addition, using a conceptual model to guide the development simpli es the
acquisition of the genetic data and can be precisely linked to the bibliography.
However, the study of the prototype performance working with real DNA
samples must be analyzed. In order to ful ll this task, further studies related
with sequencing algorithms and tools will be carried out.
Conceptual modeling of genes is not a completely novel research area. Some
works [14] [15] [16] to organize the genomic data have also been proposed be-
fore. The main contribution of the presented work is that the conceptual model
proposed here is speci cally designed to guide the implementation of software
artifacts using a model-driven development approach.
As further work it is planned to extend the GIS prototype with the aim of
achieving a higher accuracy and to facilitate the input of sequences. As a nal
goal, the GIS prototype will be tested in a real environment by means of a
collaboration with IMEGEN, a genomic medicine institute, and a couple of local
hospitals.
Acknowledgments
This research work has been developed with the Generalitat Valenciana support
under the project ORCA (PROMETEO/2009/015)
References
1. Gilbert, D.G.: Eugenes: a eukaryote genome information system. Nucleic Acids
Research 30 (2002) 145148
2. Navigenics. http://www.navigenics.com (2010)
3. 23andMe. https://www.23andme.com (2010)
4. deCODEme. http://www.decodeme.com (2010)
5. Irizarry, R.A., Bolstad, B.M., Collin, F., Cope, L.M., Hobbs, B., Speed, T.P.: Sum-
maries of aymetrix genechip probe level data. Nucleic Acids Research 31 (2003)
e15
6. Klein, R.: Power analysis for genome-wide association studies. BMC Genetics 8
(2007) 58
7. den Dunnen, J.T., Antonarakis, E.: Nomenclature for the description of human
sequence variations. Human Genetics 109 (2001) 121124
8. Richesson, R., Turley, J.P.: Conceptual models: De nitions, construction, and
applications in public health surveillance. Journal of Urban Health 80 (2003) i128
9. Pastor, O., Levin, A.M., Casamayor, J.C., Celma, M., Villanueva, M.J., Eraso,
L.E., Alonso, M.P. Enforcing conceptual modeling to improve the understanding
of human genome. Research Challenges in Information Science (RCIS 2010)
10. NCBI: The RefSeqGene project. http://www.ncbi.nlm.nih.gov/RefSeq/RSG
(2010)
11. Stevens, R., Goble, C., Baker, P., Brass, A.: A classi cation of tasks in bioinfor-
matics. Bioinformatics 17 (2001) 180188
12. Altschul, S., Gish, W., Miller, W., Myers, E.W., Lipman, D.: Basic local alignment
search tool. Journal of Molecular Biology 215 (1990) 403410
13. Gene Codes Corporation.: Sequencher. http://www.genecodes.com (2010)
14. Consortium, T.G.O.: Gene ontology: tool for the uni cation of biology. Nature
genetics 25 (2000) 2529
15. Paton, N.W., Khan, S.A., Hayes, A., Moussouni, F., Brass, A., Eilbeck, K., Goble,
C.A., Hubbard, S.J., Oliver, S.G.: Conceptual modelling of genomic information.
Bioinformatics 16 (2000) 548557
16. Ram, S.: Toward Semantic Interoperability of Heterogeneous Biological Data
Sources. In: Advanced Information Systems Engineering. Springer Berlin / Hei-
delberg (2005) 32