I have previously been involved in a course, where we (the students) were supposed to use a specific style, when presenting our graphical data. By style I mean the size/color of legends, thickness of datapoints, layout of the figure, etc....
Does similar guidelines exist when publishing data in a peer-reviewed journal? Or is it entirely up to me to decide, how it should look?
The purpose of any style when presenting graphics should aim at making the information as easy to read and comprehend as possible. I do not think any journals would provide specific style recommendaions except provide information on, for example, the thinnest posible lines, smallest possible font sizes to be reproduced in printing, color models etc.
I think it is important to think about how you present data in any scientific or professional communication. It will be largely up to you to make decisions on the material you visualize. My recommendation is to look at the book The Visual Display Of Quantitative Information by Edward Tufte and then to learn by example. This book, in my opinion, like no other, provides a good basis for understanding graphic display of data. Look at what others do and critically evaluate how successful those attempts are.
There may be journal-specific guidelines, but none of the journals I dabble in have specified such. I tell students that journal articles require "professional" looking graphs. This usually translates into simple graphs. If you take something like the Excel defaults you will see the opposite of professional looking. Excel puts lots of colors, too small fonts, lines everywhere, titles, unlabeled axes, etc. Simple black and white plots work just fine for most plots. Your job is to present the data for easy access to the reader. If you want to understand this, just grab some recent journal articles in top journals and look at their graphs. You'll see what good graphs should look like.
I'd suggest just using a good graphics program and using its defaults. This would normally look good enough, I imagine most journals would not be too fussy about such things. You probably also want to use a scripting language for reproducibility and efficiency. The following packages are reasonable choices, produce high quality graphics without much work, have reasonable defaults which can be tweaked if necessary, and export into the common image file formats like PDF, PNG etc. ggplot2 in particular is designed to be very high level and flexible, and is based somewhat on the Grammar of Graphics book by Wilkinson, which describes a high level approach to data visualization. Specifically, high level implies ignoring the small details of your graphic and letting the program taking care of it for you.
The R package ggplot2 already mentioned is quite popular. You could also try PGF/TikZ and the plot package that uses it, pgfplots. matplotlib is another possibility, though I have not used it much. Also note that both R and matplotlib have PGF/TikZbackends. For R see TikZDevice though it does not seem to be actively developed. For matplotlib, see matplotlib: Typesetting With XeLaTeX/LuaLaTeX.
