How can I adjust the threshold in TreeBagger and other related prediction methods?

From my understanding, when a method such as TreeBagger estimates probabilities of falling into a class, the probability is calculated as the number of times out of B trees that the observation is classified into each class. For example, if observation i is classified into class K in seventy out of B = 100 trees, then the estimated probability of observation i belonging to class K is .70. The classification at the individual tree level is my concern. Again from my understanding, at the individual tree level an observation is classified as belonging to class K if the node to which the observation belongs contains the the highest proportion of observations from class K. If this was a binary outcome then if the number of observations in the terminal node contain more than 50% from class K all observations in that node are classified as coming from class K. How can this threshold be changed? In other words, is it possible to classify all observations from a given node into class K if the node contains p% from class K, where p<.5?