Test selection help for multivariate response and predictors

[jst04004]

I need help determining what method would be applicable to the following situation.
There are 2 chemicals, A and B. Also, there are two organisms, say Jim and Kelly.
It is impossible to separate Jim and Kelly. Although, sometimes, levels of Jims or Kellys might be slightly more than the other. When Jim and Kelly are together, the levels of A and B decrease, but it is unknown if it is Jims or Kellys that are responsible, or possibly if it is an interaction, or if say Jims work on A, and Kellys work on B, vice versa, or if there is an interaction of effects.
It is possible to measure levels of Jims and Kellys, and to measure levels of change of A and B.

I thought that a MANOVA would work, but with sampling the variables are more on a continuous scale than categorical, and it would be difficult to draw lines in the sand for categorization. I haven't done stats in a while, and was wondering if anyone out there had a suggested test, regression method, etc. to help determine which organism is responsible for what. Any help would be greatly appreciated, and I'll buy you a beer if I ever meet you.

[hlsmith]

Think in the same terms and actions as MANOVA, but try applying generalized linear modeling. Search GLM and see if it may be a good fit.