Hello and thanks for your interest. Paired-end reads tend to be better just because they generally have more uniquely mapped and splice junction reads (the only ones MAJIQ considers). However, MAJIQ can handle single-end reads just fine as well.
Other other thing to consider. Sequencing depth is very important where more is better. What sort of depth is your 15 samples? Typically the lower bounds is something like 20M reads. Finally if your library is strand specific (forward or reverse), MAJIQ can account for this as well. This information is given to the MAJIQ Build command through the configuration file. More information on that is in the documentation on the configuration files here: https://biociphers.bitbucket.io/majiq/quick.html#conf_file
Let me know if you have any more questions!