Researchers at Johns Hopkins University have developed a web-based application, dubbed Pyromaker, to help researchers interpret complex mutations in pyrosequencing-based diagnostic assays.
Pyromaker generates virtual pyrograms — or simulations of the peak patterns that would be generated by a pyrosequencing system — for a given mutation based on information users enter into the system, such as the percentage of tumor and normal cells, the wild-type sequence, the dispensation order, and any number of mutant sequences.
The program calculates a "virtual trace" of the expected pyrogram, which can then be compared visually to actual ones generated by sequencing experiments to see if the predicted outcome matches reality.
The program was developed to help users clarify ambiguous results in diagnostic assay data. Specifically, it can catch complex mutations — such as cases where two adjacent bases are mutated — that might otherwise go unnoticed, the researchers said.
James Eshleman, associate director of the molecular diagnostics laboratory at the Johns Hopkins University School of Medicine and the paper's senior author, told BioInform that Hopkins is using Pyromaker to resolve complex BRAF and KRAS clinical results.
Eshleman's team has also developed a teaching module for the software meant to help users explore the impacts of varying parameters.
He pointed out that using Pyromaker is a cheaper alternative to additional tests that would otherwise be required to make sense of ambiguous results.
Usually, researchers apply techniques like TA cloning, Sanger sequencing, and melting curve analysis to make sense of complex mutations, he noted.
However, the authors note that although these methods provide "unequivocal interpretation," they are labor intensive, pose the risk of plasmid contamination, may delay reporting, and are not routinely used in most clinical diagnostic laboratories.
The authors note in the paper that Pyromaker is just as effective as these other methods, but because it is free, it is the "least expensive and fastest method to resolve these cases."
To run Pyromaker, users can select one of two modes. In the first one, dubbed the hypothesis testing mode, users enter alternative hypotheses for an ambiguous reading, as in when it is difficult to tell whether observed mutations are on the same allele or two different alleles. In this example, users would enter both hypotheses using Sanger sequencing data and see which pyrogram matches actual data from the clinical experiment, Eshleman explained.
In the second mode, the user iteratively adds in mutations to the wild-type pyrogram until the results match those generated by the pyrosequencing experiment, the researchers explain in the paper. The benefit of this mode is that it can be done with just the experimental pyrogram and the wild-type data; no other information is required.
In the paper, the researchers report that both modes were able to successfully identify complex mutations in codons 12 of the KRAS gene that matched the results from clinical experiments.
The authors used Qiagen's PyroMark Q24 pyrosequencing system and the company's KRAS v 2.0 test in their analysis.
Compared to some other software programs that automate the interpretation of mutations, Pyromaker "doesn’t try to interpret [mutations] fully," Eshleman said. It just "produces the expected results ... [and] then it's up to a human to compare the patterns to see if it correctly fits the actual result."
In the paper, the authors note that many labs use Qiagen’s KRAS pyrosequencing interpretation software, called KRAS plug-in report, but this software was "unable to correctly identify the mutations" in the two complex codon 12 cases.
Eshleman said that Pyromaker could be tweaked for use with short-read sequence data, but that’s not necessary because such sequencers read complex mutations as contaminants and solve that problem by simply throwing these variants out of the final output.
"The reason for that is usually you have hundreds of thousands or millions or tens of millions of reads, so even if you throw out five percent of them, you have more than enough" to interpret the data, he said. "In [our] case we have to deal with it since its one sample from one patient with either lung cancer or colon cancer and we have to come up with an answer."
Have topics you'd like to see covered in BioInform? Contact the editor at uthomas [at] genomeweb [.] com.