Data processing: database and file management or data structures – Database design – Data structure types
Reexamination Certificate
2000-09-28
2004-05-25
Metjahic, Safet (Department: 2171)
Data processing: database and file management or data structures
Database design
Data structure types
C707S793000, C707S793000
Reexamination Certificate
active
06741983
ABSTRACT:
TECHNICAL FIELD OF THE INVENTION
The invention is related to the area of computer software for the management of databases. In particular it is related to the field of tree-structured indexing methods for the rapid storage and retrieval of DNA profile information from databases containing a large number of records.
BACKGROUND OF THE INVENTION
Existing database indexing methods exploit the structure inherent when more than one database field is used. These methods are commonly based upon space-filling curves to map the multi-dimensional data to a single dimension, which is then indexed in the standard fashion. The B-tree indexing algorithm [1] and similar algorithms attempt to maintain a balanced index tree by adjusting the thresholds used to split the indexed parameter's value set as the tree is descended. Multi-dimensional indexing methods are found under several names, such as R-trees [2] and R*-trees [3], and applications exist in the implementation of image databases and other areas. A parallel database based upon this type of approach has been patented by IBM [4] using MPI, a widely available message-passing interface library for parallel computing [5]. Other implementations exist in some commercial database systems, such as the Informix Dynamic Server's Universal Data Option [6].
DNA profile information consists of allele information at one or more DNA loci or sites. Typically 10 or more loci are used. Typically, individuals can exhibit either one or two alleles at each site; forensic samples containing DNA from two or more individuals can have more alleles. The anticipated size of databases containing DNA profile information necessitates new methods to manage and utilize the stored information. An example of such a database is the national CODIS [11 ] database, which is expected to eventually store on the order of 10
8
profiles and uses complex match specifications. Standard database indexing structures such as B-trees, which provide rapid access to records based upon the value of a selected database field, are not able to take advantage of naturally occurring structure in the data. Although more than one field may be indexed, the index structures are computed independently. Retrieval of stored information based upon several indices requires an intersection of the results of retrievals based upon each index, which is a time-consuming operation. Methods using R-trees, R*-trees, and similar approaches rely on space filling curves rather than structural properties of the data. There remains a need in the art for database structures and search engines that can rapidly and efficiently store, manage, and retrieve information from very large datasets based upon the structural properties of the data expressed in multiple fields.
SUMMARY OF THE INVENTION
By way of example and without limiting the application of the present invention, it is an object of the invention to organize the storage of DNA profile information to minimize the time required to locate all DNA profiles within the database that satisfy a set of user-selected criteria when compared against a target DNA profile and therefore match the target.
It is a further object of the invention to provide a method for the parallel implementation of a database of DNA profiles by breaking up the work involved in storage and retrieval of sets of information into many requests for work which may be distributed among a cooperating group of computer hosts to balance the workload across the hosts and thereby minimize the time required to perform the work.
These and other objects of the invention are provided by one or more of the embodiments described below.
One embodiment is a method for performing a retrieval operation in a database comprising a tree of nodes. The tree of nodes comprises a root node which is connected to two or more branches originating at the root node. Each branch terminates at a node. Each node other than the root node may be a non-terminal node or a leaf node. Each non-terminal node is connected to two or more branches originating at the non-terminal node and terminating at a node. Each leaf node comprises one or more data records of the database. A test is associated with each non-terminal node that defines a partition of data records based upon either entropy/adjacency partition assignment or data clustering using multivariate statistical analysis. A current node is initially set to the root node. Input is received of a search request providing a retrieval operation and information necessary to perform the retrieval operation. The test associated with a current node is performed responsive to the search request. The test results in identification of zero or more distal nodes connected to the current node. The identified distal nodes can, according to the test, contain the data record. The test is repeated using an untested distal node which is a non-terminal node as the current node. The retrieval operation is performed on each referenced node that is a leaf node.
Another embodiment is a method of partitioning data records in a computer into groups of roughly equal size. A function is defined of the probability distribution of the values of a designated variable associated with the data records. The function comprises a linear combination of measures of entropy and adjacency. The values of the designated variable are partitioned into two or more groups such that the value of the function is minimized. Each data record is assigned to a group according to the value of the designated variable.
Yet another embodiment is a method of creating a tree-structured index for a database in a computer. The database comprises a tree of nodes. The tree of nodes comprises a root node which is connected to two or more branches originating at the root node. Each branch terminates at a node. Each node other than the root node may be a non-terminal node or a leaf node. Each non-terminal node is connected to two or more branches originating at the non-terminal node and terminating at a node. Each leaf node comprises one or more data records of the database. The tree-structured index comprises one or more tests associated with each non-terminal node. Naturally occurring sets of clusters are identified in the data records of the database. For each identified set of clusters, a test is defined that assigns each data record to a cluster within the set of clusters. Each such test is associated with a non-terminal node, together with an associated set of clusters. One branch is associated with each cluster within the set of clusters. The branch originates at the non-terminal node and forms part of one or more paths leading to leaf nodes comprising the data records assigned to the cluster by the test.
Still another embodiment is a method of organizing the data records of a database into clusters. One or more variables in each data record are represented in a binary form, wherein the value of each bit is assigned based on the value of a variable. A set of variables is chosen from those represented in all of the data records such that principal component analysis of the set of variables yields distinct clusters of the data records. Principal component analysis is applied to a sample of the data records, and two or more principal component vectors are identified, whereby the scores of the sample data records along these vectors form distinct clusters. A test is formulated based on the identified principal component vectors which assigns each data record to a cluster. The test is then performed on each data record, and the data records are organized into clusters.
Another embodiment is a parallel data processing architecture for search, storage, and retrieval of data responsive to queries. The architecture includes a root host processor that is responsive to client queries; the root host processor creates a search client object and establishes an initial search queue for a query. The architecture also includes a plurality of host processors accessible by the
Birdwell John D.
Horn Roger D.
Icove David J.
Wang Tse-Wei
Yadav Puneet
Banner & Witcoff , Ltd.
Metjahic Safet
Nguyen Merilyn
LandOfFree
Method of indexed storage and retrieval of multidimensional... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of indexed storage and retrieval of multidimensional..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of indexed storage and retrieval of multidimensional... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3243743