Weights of the variables must be positive

[llrs]

I don't know if mathematically is reasonable but the fact that a contribution of a variable can be negative is strange.

For instance a gene might be needed to go down, but still its contribution (sgcca_result$a$gene[1, 1]) is (or should be) positive. Otherwise the interpretation of the weight is difficult because if you take it as an absolute value you are saying that it is as important to go up than to go down.
Test something on these lines, probably with sequential implementation too. (forcing all the weights in a to be positive via abs)