Hi everyone , I will have to work on an exam and the topic is to build a Visual Analytic Application about any dataset we want. Im new in this field (BioInformatic) but if I have to spend a month in deep programming I wish I could help this community somehow. Im doing Thesis on the role of platelets in cancer and so I know how to obtain differentially expressed genes and everything related to it. Now the project have some rules of which :
Every assignment MUST have a visual part constituted by at least 2 Visualizations coordinated in both ways and interactive.
ANALYTICS: every assignment MUST contain at least 1 computation that is triggered by user visual interactions
Mandatory: every assignment MUST use a PCA or a MDS or a t-SNE algorithm.
Now I though about using a count matrix with sample informations. My idea was to project a PCA analysis over the samples and DE genes, from which you can select directly on the plot the samples you like to study more triggering a second visualization (Heatmap) that would change based on the selection. Furthermore I would additionally add a filter over the genes based on FDR and logFC values previously computed using one of the packages between edgeR and Deseq2. Maybe adding some scatterplot or volcano plot to give even an easier selection of genes. Still have to think about an Analytic to add.
Please if you find that PCA is not applicable or stupid dont mind to tell me, but would be even better if you can explain me what would be better in this case. I'm new in this ( I come from Software Engineering ) but I really wanna help this community (I know sounds dumb but I kinda feel more purpose working in this field) so please feel free to propose what so ever and to critic my choices.
Sounds like you try to reinvent existing applications like Glimma, iSEE, pcaExplorer or interactiveComplexHeatmap and others. Be sure to extensively go through literature and do a solid google search, these types of applications you describe already exist. So unless you are bringing in something fundamentally new it makes limited sense to reinvent said wheel.
Would it be useful to maybe expand an analysis over Differential co-expression (DCE) analysis? Maybe being able to interact with it seeing for example the genes that have strong degree of connection helping a future classification?
It's bioinformatics with an s. You're not the first well intentioned programmer that thinks they can invest a fraction of the time that people with domain knowledge do to solve a widespread problem with new fangled tech. I think that's a mandatory phase in the career of any CS person that turns to bioinformatics. At this point, I'm glad you haven't mentioned "cloud" and "workflow" in your post :-)
Sorry i did not get the comment, what do you mean for "didn't mention cloud and workflow". As I said I'm new here (and the s on BioInformatics is because in Italy we call it BioInformatica as singular , so i confused myself) and Im not trying to be a hero for sure, is a project for an exam from which it could just come out something interesting, thats it. Just wanted to ask some tips :).
It's kind of weird that you say you wish to help this community when it looks like you actually need the community's help for your exam project. If it's for your learning purposes, reinventing the wheel is not too bad - an interactive volcano that lets you pick and gene and then look at pathways using GSEA would be nice but please be clear - this is more a project for you than addressing the community's need for something specific. If the community had a requirement that could be boiled down to something that can be explained to a new person, that problem has probably already been solved.
Yes this is indeed a question for my project , I said that , not hiding it at all. Im just asking if there is something not done that I could add (this is not mandatory in my project). And about the : " you wish to help this community when it looks like you actually need the community's help for your exam project " is totally wrong. Im literally basing my master thesis in cancer prediction because I actually have a disease and as I would love that someone would study mine I decided to apply for someone else even if I started from zero. So if you wanna play like "You can not understand how it feels to work in BioInformatics" , then thanks for your comment but I don't need it.
ali, I'm a CS person that moved to bioinformatics. I struggled for nearly a year to implement basic CS stuff at my first job. Trust me, I'm telling you that I've been where you are. You wish to help and your ideas are fascinating but people have bigger problems that CS skills alone cannot solve.
Anyway, good luck! I hope you find your niche and are able to contribute to the community.