1. ABSTRACT VRLGS, or Virtual Restriction Landmark Genome Scanning, is a program that allows a user to virtually generate the gels that are obtained by the RLGS technique in a genetics lab. In the lab setting, RLGS generates large two-dimensional gels using a DNA gel electrophoresis technique that can be used to compare DNA patterns and identify changes in normal versus cancerous tissues. The creation of such gels in a lab setting is not easy due to the time-intensive and expensive nature of these gels. As a result, VRLGS was created in an effort to help alleviate these problems through the creation of Virtual RLGS gels. It allows for an array of features to be offered, including user-defined algorithms in the computation of DNA migration and spot generation. Unlike its lab counterpart RLGS, VRLGS also allows for the retention of vital spot information, such as chromosomal locations. Furthermore, VRLGS can also be used to virtually generate gels that have only specific sequences of interest (i.e. virtual running gels) in order to elucidate important spot information. As a result, VRLGS significantly addresses some of the problems posed by traditional RLGS offers some advantages over the traditional lab- based technique, and when used in conjunction with RLGS, can be a very powerful tool. RLGS RLGS, an abbreviation for Restriction Landmark Genome Scanning, is a two-dimensional DNA gel electrophoresis technique that can be used to compare DNA patterns. Although it is often used in genome mapping projects, RLGS can also be combined with restriction enzymes that are methylation sensitive to help detect differences in methylation patterns at CpG dinucleotides within CpG islands. Dr. Plass has used RLGS to compare gels of normal tissue DNA to that of tumor tissue DNA to identify alterations in CpG island methylation, which can be characteristic of various types of cancer. These alterations may affect gene regulation by either turning on tumor growth promoting genes or by turning off tumor growth suppressing genes. Upon identification of CpG island methylation, the lab can extract information from the RLGS gel and identify clones of these sequences in their library, which can then be used to study the effects of the sequences in cancer. In the past, Dr. Plass has created gels by digesting sequences with three restriction enzymes, NotI, EcoRV and HinfI, and then running the gels using the RLGS technique. Now, we are exploring the possibilities of virtually creating such gels. VRLGS: Virtual Restriction Landmark Genome Scanning VRLGS VRLGS, or Virtual Restriction Landmark Genome Scanning, is a program that allows a user to virtually generate the gels that are obtained by the RLGS technique in a genetics lab. This program takes a standard fasta file and then through a series of scripts creates a TIFF image. The computation of the migration of the DNA sequences is accomplished by providing the user a choice of various algorithms including ones that take into account DNA curvature. Furthermore, the program also allows the user to define the algorithm used to generate the spots. VRLGS can be used to virtually create gels that, in the lab, are large and complicated to run, such as a gel that has spot information for all of the chromosomes in the human genome. Furthermore, upon generation, these gels are still able to retain vital spot information, such as the chromosomal locations. VRLGS can also be used to virtually generate gels that have only specific sequences of interest (i.e. virtual running gels) in order to elucidate important spot information. VRLGS has three separate elements in its development. The first involves mimicking the fragmentation of the DNA sequences with the methylation-sensitive enzymes. The second is the prediction of the spot locations within the image and the migration of the DNA sequences. The final step focuses on the generation of the image to interface with Conime. Advantages VRLGS allows for the virtual generation of these RLGS gels and offers many advantages over RLGS: Less Expensive – In RLGS the gels require the use of restriction enzymes to fragment the DNA, which are quite expensive. Furthermore, the developing process requires the use of expensive chemicals, including radioactive materials. VRLGS, on the other hand, bypasses all of these materials and creates virtual gels. Easily Create Numerous Gels – In RLGS, several samples of tissue, both normal and cancerous, need to be obtained and have gels ran on them. In VRLGS, tissue collection is not necessary for gel creation. Time-Intensive – RLGS gels take a tremendous amount of lab time to generate (often about a week) and may not always develop into a usable gel. VRLGS gels take at most, a few minutes to generate. Limited Number of Restriction Enzymes – RLGS gels are limited in the number of restriction enzymes that can be used in their creation. This limitation does not provide researchers with the opportunity to study and identify all the various sequences that are involved in cancer. Using VRLGS, we are able to test various enzyme combinations and identify the best combinations to use, in order to ensure the least redundant analysis of genomic loci. Chromosomal Gels – Through VRLGS, a gel representing vital chromosomal information, such as fragment length and chromosomal location, can be generated. This allows for the crucial sequence identification of RLGS spots that exhibit changes in lab-based gels. DNA Migration – Prediction Another problem that was encountered in the development of VRLGS was determining an algorithm that could effectively simulate the migration of the DNA sequences in a gel. The DNA fragments migrate first in the x-direction, and then migrate in the y-direction. Smaller fragments travel farther due to their easier passage through the porous gel matrix, however, DNA curvature also has an impact, which seems to be dependent upon the sequence. The following algorithms were used to simulate this effect: Length – Originally a length-based algorithm was used. The algorithm was generated by taking known information about spots, including their sizes, in base pairs, and the location of migration their within the gel, and using it to come up with a formula. This prediction system was not quite accurate, because the smaller fragment would move farther both in the x- and the y-directions creating a natural curve. Curvature – Three different curvature algorithms were applied. Originally, the evidence that was used to generate the length-based formula, was manipulated in Matlab and curvature was taken into account. That formula was then manipulated, by taking into account greater spot evidence to generate a better predictor. Shading One major obstacle involved in the development of VRLGS was developing an algorithm that effectively emulates the natural shading exhibited by the spots in the gels. The shading drops off naturally, due to the inherent biological nature of the gels. The following algorithms were used to help implement natural shading: Distance – Originally, a distance-based algorithm was used, whereby the pixel shading would drop down by some factor, the farther away the pixel was from the center. This wasn’t very effective, because it gave the spots a very choppy, pixelated appearance. Gamma Correction – a logarithmic factor that can take a linear algorithm and apply a correction factor, by which to correct it logarithmically. The correction factor was user defined. Although it helped a little, the pixelation was still very apparent in the appearance of the spots. Gaussian Curvature – A mathematical/statistical algorithm that accounts for curvature using the following formula: This formula was best at representing the curvature that is inherently found in RLGS. Curvature Length Computer & Information Science Dept. The Ohio State University 395 Dreese Labs 2015 Neil Avenue Columbus, OH 43210 Human Cancer Genetics The Ohio State University 494 Tzagournis Laboratory 480 12 th Avenue Columbus, OH 43210 Biomedical Informatics The Ohio State University 3184 Graves Hall 333 West 10th Avenue Columbus, OH 43210 Computer and Information Science DepartmentHuman Cancer GeneticsBiomedical Informatics Dr. Rephael WengerDr. Christoph PlassDr. Ilya Ioschikhes Ramakrishnan Kazhiyur-MannarDr. Dominic Smiraglia Yisheng Chen Tahmina Ansari