Frequently Asked Questions about the Bioinformatics/Neuroinformatics module.
Q1: What if the websites involved in the module go down?
Q2: When using the Entrez Gene website, can other reference assemblies be used?
Q3: My students all seem to get slightly different numbers and slightly different distributions when they run the permutation test on the GeneNetwork (formerly WebQTL) website. Is this normal?
Q4: My students found that when they controlled for brain weight while running the multiple regression, body weight (which previously accounted for a significant portion of the variance) was no longer significant. Is this possible?
Q5: Is it possible for the residual variance to actually go up when new variables are added to the multiple correlation?
Q6: When I tried to open images from the All Mice v1 – Image Library (all of which were downloaded from the Mouse Brain Library website), my computer choked and refused to open the file. Why does this happen?
Q1: What if the websites involved in the module go down?
Teaching with web-based tools does have this risk. The GeneNetwork (formerly WebQTL) and UCSC Genome Browser websites have both been occasionally inaccessible during teaching. These were rare occurrences, and there was never an instance in which both sites were simultaneously down. As a back-up, instructors should pre-run the data and take screenshots of the output to present in such cases.
Q2: When using the Entrez Gene website, can other reference assemblies be used?
Yes. GenBank (http://www.ncbi.nlm.nih.gov/Genbank) is also used.
Q3: My students all seem to get slightly different numbers and slightly different distributions when they run the permutation test on the GeneNetwork (formerly WebQTL) website. Is this normal?
Yes. Every student will get a slightly different “draw” that randomly matches the markers to phenotypes when constructing this null hypothesis distribution.
Q4: My students found that when they controlled for brain weight while running the multiple regression, body weight (which previously accounted for a significant portion of the variance) was no longer significant. Is this possible?
When removing the variance due these variables sequentially, a given variable may be associated with a highly significant proportion of the variance in one analysis and then become non-significant if another variable is added. This is because the two variables are probably highly correlated (which can be checked) and one of the variables “soaks up” all of the variance common to both, leaving almost none to be strictly associated with the other.
Q5: Is it possible for the residual variance to actually go up when new variables are added to the multiple correlation?
To Yes, this can happen if the predictor variable is uncorrelated with the dependent variable of interest. Also, if two variables are highly correlated and one variable “soaks up” all of the common variance, leaving one variable to only contribute error to the prediction (see answer from Q4), then adding a variable can actually cause the residual variance to go up a little.
Q6: When I tried to open images from the All Mice v1 – Image Library folder (all of which were downloaded from the Mouse Brain Library website), my computer choked and refused to open the file. Why does this happen?
Some of image files are large enough that older computers may not be able to handle them. To get around this, open the files in Adobe Photoshop (or a similar image-manipulating program), convert them to grayscale if they are in color, and re-save them as tiff files. Image J will then have an easier time with these files. Scion Image is not recommended because it cannot handle the particle counting on large file sizes (at least on Mac).