In Bioinformatics there is a lack of software tools that t with the requirements demanded by biologists. For instance when a DNA sample 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 research path. One of the most promising approaches is the use of conceptual models in order to precisely dene 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 analysis prototype has been implemented using this conceptual model as a development core.
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 understanding of the relationship between genotype (gene conguration and combination of a particular individual) and phenotype (expression of the genes in a specic 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 dened to support an specic research area or experiment. Therefore, when biologists want to use them for a particular assay, it is very unlikely that they support their specic 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 principles, 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 dicult because it is spread around different databases.
Since several tools are required to analyze the data, the specication of the tooling workow 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 [
Following that idea, this paper presents a GIS prototype that analyzes a DNA sequence in order to nd documented variations for a specic 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 specic 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 specic 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 specic 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
In recent years, several commercial tools have been developed to provide genomic
analysis. These tools can perform tests in order to estimate the customer
probability to suer certain diseases. Navigenics [
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 classied. 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.
The main objective of the conceptual model presented in this paper is to establish a connection point between the genomic eld and the GIS development domain. One of the main characteristics of the genomic eld is heterogeneity. The unication of the relevant concepts is a dicult task, since genomic concepts are not precisely dened. Moreover, the eld knowledge is still developing and these concepts are constantly evolving, making the organization of all the genetic data available more dicult.
Genetic databases are thus aected by this heterogeneity problem. Each
database reects the concepts according to the interpretation and terminology
of a biologist. However, there are dierent denitions for the same concept; for
example, a variation in the DNA sequence is referred under the terms: variation,
mutation, polymorphism or SNP [
The conceptual model presented here claims to be precise with genetic
concepts 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 [
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 dene a variation that is completely located and Imprecise variations, whose location details are not specied. 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 specied in the model is always related to its phenotype, which is modeled by the Phenotype entity (3). The Certainty entity species the probability that a phenotype could show up because of a concrete variation on the genotype. In case is identied 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 dened by the Bibliography 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
The main goal of the prototype is to show how conceptual models can be useful
to dene a GIS. One of the most common tasks in the genomic area is the
analysis of the genomic sequences [
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 dened conceptual entities. Thanks to this conceptualization, heterogeneity and data dispersion problems are solved, avoiding the manual preprocess of some noncomputer legible data and ensuring the quality of the data stored.
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
prototype, 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
variations. This alignment is performed using the BLAST algorithm [
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 denes formally the concepts related to genomic variations. As a proof of concept, a GIS prototype, which uses this conceptual model as background, has been implemented.
One of the advantages of using the presented GIS prototype is that the variation analysis can be performed using only one tool, avoiding the data workow. In addition, using a conceptual model to guide the development simplies 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 fulll 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 [
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)