0
$\begingroup$

I have no expertise in biology, I'm a data scientist, and I would like to know if it makes sense, from the biological point of view, to analyze data (SNP data) coming from a single chromosome, and not all 22 chromosomes, to predict the risk of a certain disease.

Should I obligatorily use data from all chromosomes? Why?

Thank you very much. And sorry if it is a very basic question, but I really would like to understand this.

Improve this question
$\endgroup$

2 Answers 2

Reset to default
2
$\begingroup$

You can analyse a single chromosome. All the bioinformatics tools I know, allow that. Consider only the reads that map to the desired chromosome (or any segment of the genome). However, most next-gen sequencing experiments sequence the whole genome and not just a single chromosome. You'll just end up losing information by not considering other chromosomes.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Ok, understood. So you are sying that it finally depends on the final objective of the analysis, right? In my case, I want to analyse the data (dataset 1: lung cancer individuals, and dataset 2: type 2 diabetes individuals) to create a final prediction model. Also the idea is to perform feature selection and detect relevant/significant SNPs to later identify related genes and see if they are related with the disease or not. So the ultimate question would be, for complex diseases like lung cancer and type 2 diabetes, should I analyse data from all chromosomes or only some? $\endgroup$
    – mgvaldes
    Commented Jan 23, 2017 at 10:33
  • $\begingroup$ Even though tools let the individual analysis of a single chromosome, is it biologically correct? Does it make sense from a biological point of view? Does it depend on the disease being analyzed? I'm asking because biologists have told me that the conclusions of the analysis I'm doing, can olny come from the analysis of data from all chromosomes, but I would like to know the biological justification of this. Hope I made my self clear. Thank you! $\endgroup$
    – mgvaldes
    Commented Jan 23, 2017 at 10:45
  • 1
    $\begingroup$ @mgvaldes it would not be a good idea to restrict yourself to one chromosome unless you have a good reason to do so (for example, X-linked genes). Biological justification for this is that complex traits are dependent on many genes that are scattered across different chromosomes. $\endgroup$
    – WYSIWYG
    Commented Jan 23, 2017 at 10:49
  • $\begingroup$ Perfect. That's what I wanted to be sure of, since I'm not a biology expert. Thank you very much! $\endgroup$
    – mgvaldes
    Commented Jan 23, 2017 at 10:51
0
$\begingroup$

On top of @WYSIWYG's answer, suppose you want to analyze production of certain protein A produced by gene A on some chromosome A. Your data will be like say, luminosity/fluorescence values coming out of tagging this specific protein. But to predict this effect of production (on what period this values might be higher, like in a prediction model), the gene/factor that might be activating gene A might be on some other chromosome B. Until you know the activity of that, you will not be able to accurately predict the production of this.

In a Regression/Tree model, your $X's$ will be categorical/continuous showing the activity of effecting genes(B) while activity of gene A is dependent $Y$.

Hope this helps.

Improve this answer
$\endgroup$
1
  • $\begingroup$ I'm concretly applying ML learning techniques to SNP data from all 22 chromosomes of two separate types of individuals (dataset 1: lung cancer, dataset 2: type 2 diabetes). The final idea is to create a predictive model for each use case and rely on feature selection techniques to identify relevant SNPs realted to the disease. Read the other comments I've given to @WYSIWYG's answer. $\endgroup$
    – mgvaldes
    Commented Jan 23, 2017 at 10:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .