Method and apparatus for comining gene predictions using bayesian networks

Abstract

Computer based apparatus and method automates gene prediction in a subject genomic sequence. A plurality of expert systems provide preliminary or intermediate gene predictions. A Bayesian network combiner combines the intermediate gene predictions and forms a final gene prediction. The final gene prediction accounts for dependencies between individual expert systems and dependencies between adjacent parts of the subject genomic sequence.

1 . A computer apparatus for gene prediction comprising: 
 a processor that is adapted to execute instructions that cause the processor to 
 create a plurality of units that predict gene locations in a subject genomic sequence, wherein each of said units is capable of providing respective intermediate indications of gene locations output; and  
 create a combiner that receives said respective intermediate output indications of predicted gene locations, the combiner comprising a Bayesian network that combines said intermediate indications of gene locations using probabilities of gene locations of the subject genomic sequence to form a final combined output for indicating predicted gene locations in the subject genomic sequence.  
   
     
     
         2 . The computer apparatus as claimed in  claim 1  wherein the plurality of units is a plurality of expert systems.  
     
     
         3 . The computer apparatus as claimed in  claim 1  wherein the Bayesian network includes probabilistic dependencies between individual units and dependencies between adjacent parts of the subject genomic sequence.  
     
     
         4 . The computer apparatus as claimed in  claim 3  wherein the Bayesian network combines the predicted gene locations according to  
         Y* =max Yt   P ( Y   t   |E   1   , . . . , E   n   , Y*   t−1 )  E   i   ε{E,I}   
       where t is location in the subject genomic sequence and E 1 , . . . , E n  are the respective predictions (E for exon or I for intron) made by individual units 1 through n, n being the number of units in the plurality.  
     
     
         5 . The computer apparatus as claimed in  claim 1  wherein the subject genomic sequence is a DNA or RNA sequence.  
     
     
         6 . The computer apparatus as claimed in  claim 1  wherein gene locations include exon predictions.  
     
     
         7 . The computer apparatus as claimed in  claim 6  wherein gene locations further include exon and intron predictions; and the final combined output indicates exons and introns of the predicted genes of the subject genomic sequence.  
     
     
         8 . The computer apparatus as claimed in  claim 1  wherein the Bayesian network comprises a table or set of probabilities of a given sub-sequence being a protein encoding exon prepared by applying training data to the computer apparatus, wherein said training data comprises character strings representing known genes of a known genome sequence.  
     
     
         9 - 13 . (Canceled).  
     
     
         14 . A computer apparatus comprising: 
 means for obtaining from a plurality of expert systems a plurality of respective preliminary gene location predictions for a subject gene in a subject genomic sequence;    means for inputting into a digital processor programmed to contain a Baysesian network a plurality of respective datasets representing said gene location predictions;    means for combining said respective datasets in said Bayesian network, to form a combined Baysesian network containing probabilistic dependencies between individual expert systems and dependencies between adjacent parts of the subject genomic sequences; and    means for providing from said combined Baysesian network a data output indicating an improved predicted location for said subject gene in the subject genomic sequence.