miRDB is an online database for predicted microRNA
targets in animals. MicroRNAs are involved in many diverse
biological processes and they may potentially regulate the functions of
thousands of genes. One major issue in miRNA studies is the lack of
bioinformatics programs to accurately predict miRNA targets. Animal
miRNAs have limited sequence complementarity to their gene targets,
which makes it challenging to select relevant biological features to
build target prediction models with high specificity. We have developed
a new miRNA target prediction program based on support
vector machines (SVMs) and
high-throughput training datasets. By systematically analyzing
public high-throughput experimental data, we have identified novel features that
are important to target downregulation. These new features as
well as other known features have been
integrated in an SVM machine learning framework for the training
of our target prediction model. Our prediction algorithm has
been validated by independent experimental data for its improved
selectivity on predicting a large number of miRNA downregulated
|What is SVM?
|support vector machines (SVMs) are universal constructive
machine learning procedures based on statistical learning
theory. SVM has been applied in many diverse applications such
as pattern recognition, computational biology, and image
analysis. The basic concept is that by maximizing the separation
between the two classes in a nonlinear feature space, SVM not
only reduces the training error but more importantly also
achieves better generalization on unseen data.
interpret the target prediction score?
|All the predicted targets have target prediction scores
between 50 - 100. These scores are assigned by the new
computational target prediction algorithm. The higher the score, the more confidence we
have in this prediction. That is why the search result is
ordered by prediction score. In our experience, a predicted
target with prediction score > 80 is most likely to be real. If
the score is below 60, you need to be cautious and it is
recommended to have other supporting evidence as well.
organism is not included. What should I do?
|miRDB currently has predicted miRNA targets in five
organisms: human, mouse, rat, dog and chicken. If your organism
is not included, you may send us your request. If there is a
high demand for your organism, we will include its
predicted miRNA targets in future updates.
perform target search?
|There are two ways to search miRDB for predicted miRNA
1) Search by miRNA names. Partial names are allowed. If there
is more than one match, all the matched miRNAs will be returned
and you may choose from one of these miRNAs to view their
predicted targets. If there is only one match, the target
prediction result will be presented directly. The partial name
search can be useful if you need to do a general search. For
example, by typing in "hsa", you will retrieved all human miRNAs
with predicted targets.
2) Search by gene target information. There are three options
to do gene search: GenBank Accession, NCBI Gene ID or Gene
Symbol. You have to enter the exact ID or symbol and no partial
match is allowed. In this way, a single gene record will be
retrieved if it is predicted to be a miRNA target.
perform target mining?
|The Target Mining page provides advanced search options for
miRNAs or their gene targets. First, you need to click on one of
the radio buttons to choose either miRNA search or gene target
search. There are two optional check boxes for the exclusion of
miRNAs with too many predicted targets or targets with low
scores. You may adjust the threshold values to tailor for your
needs. The default recommendation is to only include targets
with scores >60 and miRNAs with <800 targets.
You may provide
your own miRNA or gene list. Alternatively, you may choose the
pre-compiled pathway gene list from the page. When searching for
miRNA gene targets, full mature miRNA names are required. For
the search of miRNA regulators, you may provide either NCBI gene
IDs or official gene symbols. If symbols are used in the search,
you also need to specify the species. Please use space or comma
to separate the entries.
Why the retrieved GenBank accession sometimes is different from
my submitted one?
|Because of the sequence redundancy in GenBank, each gene is
usually represented by more than one GenBank record. miRDB uses the
NCBI Gene index files to map multiple sequence records to the
same gene record. As a result, a different accession number
other than originally submitted may be retrieved. However, both
accessions represent the same gene.
|miRDB was created by
Xiaowei Wang's lab at
Department of Radiation Oncology, Washington University School of Medicine in St. Louis.