RGB 2 CMYK = WYSIN
WYG. Dont'cha love all these abbraviations?
Fundamentally, RGB and CMYK are quite different, since the former is based on adding color to black whereas the latter is based on subtracting color from white.
with a CMYK image on an RGB monitor is not optimal, since you're using an additive color display to simulate what will be a subtractive color print output. The CMYK gamut is much smaller than the RGB gamut, and also oriented differently:
When you print your RGB image at home, the print driver is converting to CMYK for you. A photo lab also converts it to CMYK, using a standard profile conversion based on the sRGB color space. In a truly color-managed workflow, where you have control over the image and the printer, you'd want to calibrate both to the same standard (which is where a tool like ColorMunki comes into play - a spectrophotometer that you use to measure the same image on your display and a print, which then generates a monitor profile to correspond to the printed output - then, WYSIWYG). Of course, with an image for publication, it's not your printer...and that makes a difference. Your printer is likely an inkjet, whereas a publication will be printed using offset lithography, i.e. halftone screens at different angles for each of the four colors.
I'd recommend converting to CMYK as a final step in the workflow, and the make adjustments based on a printed proof.
Commercial printers want you to submit files as CMYK primarily to avoid the responsibility of color shift surprises in the conversion - i.e. so if it doesn't look right to you at the proof, you
have to bear the cost of the adjustments.
In the past, the images I submitted for scientific publication needed to be CMYK (TIF or EPS). More recently, many of them have been accepting (and a couple even preferring) an RGB image with an associated ICC profile.