Week 5 Updates: ClinPrior
Hi Everyone!
This week, I finally set foot in the lab once again after a couple weeks of remote work. I’m so excited to be revisiting my site placement in person, and I have some interesting developments from this week. I enjoy being part of the lab environment. It is easier to complete work and ask for help; in addition, I heard conversations on research politics that I would not have been able to hear at home. I learned that millions of dollars worth of research grants are being withheld, yet government officials are allocating more money towards developmental research. I find it intriguing how politics affects so many branches of study, including neuroscience.
Clinprior vs MARRVEL
Over the past few days, I have worked primarily on cataloguing results from ClinPrior. I use two main files produced by the tool: the variant rank file and the overview files. The variant rank file lists gene variants by their likelihood of causing a neurological disease, while the overview file is a general list of gene variants that may be culprits in these conditions. While the rank file gives more general information on the variant, the overview file provides more specific information about the location, inheritance patterns, and deleteriousness of the variants. After matching highly-ranked variants with their corresponding overviews, I observed that the position of some variants (on the same chromosome and in the same gene) were not the same in the overview and rank files. This is likely because certain insertion or deletion mutations were accounted for in one file and not the other. Despite discrepancies between the two files, I was able to match most of the highly-ranked variants with their general overviews using the files produced by ClinPrior.
Compared to ClinPrior, AI MARRVEL encountered significantly more issues. AI Marrvel, like ClinPrior, produced both rank and overview files. While both the overview and rank files from ClinPrior listed the gene names, MARRVEL only listed the chromosome number and position. The positioning issues presented in the ClinPrior files suggested that searching for mutations by location would not be as reliable as searching for mutations by the genes themselves. In addition, MARRVEL’s deleteriousness predictions presented inaccurate results. After searching for several of MARRVEL’s highly ranked mutations in Franklin, a genomic clinical database by Genoox, I found that many of the mutations deemed deleterious by MARRVEL were likely benign.
Next Steps
For the next few days, I plan to continue working through the ClinPrior files, since AI MARRVEL is not producing promising results. I may not be doing coding for this project, but I will continue organizing the AI model’s findings. I hope to explain the results in more depth soon!
-Caitlin