How to estimate conditional probability (cdf) of multivariate dataset using "quantilePredict"?

Hi,

I used "quantilePredict" and "TreeBagger" to estimate conditional probability (CDF) for a dataset that has two features (X_1 and Y) i.e., P(X_1|Y). It works great. However, when I consider three features X_1, X_2 and Y. Then how can I find the P(X_1,X_2|Y) without the assumption of conditional independence?

How to capture the covariance as well as the CDF (quantiles)?

TreeBagger is trained (f) by giving "Y" as input and X_1 as output i.e., X_1 = f(Y). We then use the Treebagger model to predict using "quantilePredict" but in multivariate case, the Treebagger cannot fit the data where the input is "Y" and output has "X_1, X_2" i.e., Y = f(X_1,X_2) (this idea/pov is probably wrong and naive) ?.