We have used computational methods to ferret out some of this information. Using machine learning techniques we have found that 50 nt intronic stretches on either side of exons, beyond the splices site consensus sequences themselves, contain information that is necessary for the efficient splicing of most human exons (14, 12). The information at this stage is in the form of 5‑mers that are overrepresented in these regions. We would now like to know the exact nature of these signaling elements (intronic splicing enhancers, ISEs), the step(s) in splicing at which they act, and the proteins that mediate their effects.  

Additional information lies within the exon bodies in the form of exonic splicing enhancers (ESEs) and exonic splicing silencers (ESSs). Using genomic statistical analysis, we compiled lists of 8-mers as putative ESEs (PESEs) and putative ESSs (PESSs) in each class and showed that the most of the predicted motifs can function as expected (13, 11, 8). You can visit our online PESX utility to find these 8-mers in your own sequence and see reference 5 for our computational approaches.

This work along with similar successes of other laboratories now make it clear that exons and their flanks are filled with a dense population of regulatory elements. Our task now lies in figuring out how this rich sequence information is integrated to make what is usually a binary decision to splice. But the high density makes it difficult to make solitary genetic changes, and makes interpretations ambiguous.  To circumvent this problem we have turned to synthetic exons that we design to contain isolated enhancer, silencer and neutral modules (6).  We hope that the rules governing splicing will be more apparent by the pointed manipulation of these “designer exons.”

Comparative genomics is another tool we are using to decipher splicing information and to view the evolutionary pressures exerted upon these sequences. In the course of these experiments we discovered that the most recently created mammalian exons stem largely from repeated sequences and are spliced inefficiently and are often non-protein coding (10).We also find that new ESEs are constantly being created and ESSs destroyed as the genomes strives to maintain splicing efficiency in the face of continual mutation (8).

Our 2005 review dealing with the definition of splicing regulatory motifs can be found in reference 9.

A) Computational Biology
Algorithm development to predict splice sites based on information available to the cell, including secondary structure prediction, so as to identify all the elements that need to be considered for a mechanistic understanding of splicing. .

B) Designer Exons
Learning the rules of ESE, ESS, ISE and ISS interaction through the de novo design of synthetic exons.  These exons are made up of prescribed ESEs and ESSs that are designed to function singularly when combined (i.e., not give rise to confounding overlapping sequences)  

C) High throughput mutational analysis
We have recently developed this tool for carrying out saturation screens of sequences and sequence changes that affect splicing. We are using this methodology to defined combinatorial interactions governing exon definition and defining the effect of context in splicing decisions. 

D) The effect of DNA-level action on splicing
We are genetically manipulating the 25 kb dihydrofolate reductase (dhfr) locus in Chinese hamster cells to study the effect of chromatin structure and promoter/enhancer action on splicing in the natural chromosomal context of a gene.

E) A second area of interest lies in the area of biotechnology: We are isolating engineered derivatives of Chinese hamster ovary (CHO) cells that are capable of rapid gene amplification to speed up the development of recombinant protein based therapeutics, and developing vectors that increase recombinant protein production through more efficient posttranscriptional processing of recombinant transcripts.