The absorbance change of NADP => NADPH, or vice versa, is the simplest method, but the extinction coefficient change at 340 nm is only ~6200 M-1cm-1, so you need to convert at least 10 µM of substrate NADP to NADPH (or vice versa) to get a decent signal (Delta A340 = 0.06 in a 1-cm pathlength cuvette). Since you want to use up no more than 10% of your substrates, in order to maintain initial velocity, you need to start with at least 100 µM of each substrate.
For malate dehydrogenase, the reaction is (S)-malate + NADP NADPH +pyruvate + CO2. You convert one NADP to NADPH for each malate converted to pyruvate, so you need at least 100 µM each of malate and NADP.
For fatty acid synthase the overall reaction is acetyl-CoA + 7 malonyl-CoA +14 NADPH palmitate + 8 CoA + 7 CO2 +14 NADP. Therefore, to make 10 µM NADP you should start with at least 100 µM NADPH, 50 µM malonyl-CoA, and 7 µM acetyl-CoA. You also need acyl carrier protein, of course.
The other simple colorimetric assay that could be used for fatty acid synthase, in principle, is detection of the free thiol group of CoA with Ellman's reagent, which forms a product with an absorbance maximum at 412 nm and an extinction coefficient of about 14,000 M-1cm-1. It's more than twice as sensitive as the NADPH NADP absorbance change. It's normally done as an endpoint assay, however, whereas NADPH NADP detection can be done continuously. Continuous assays are nice because you don't have to quench the reaction to get time points.
A further improvement in sensitivity for CoA detection can be achieved by using Thio-Glo-1 to detect the CoA product. This is a fluorescence readout and it is about 50 times more sensitive than Ellman's reagent.
Both Ellman's reagent and Thio-Glo are incomatible with thiol-based reducing agents like DTT and mercaptoethanol.
One additional suggestion: For the malate dehydrogenase assay in which NADPH is a product, you can increase the sensitivity by over 3-fold by using the enzyme diaphorase to couple NADPH production to reduction of XTT. The absorbance increase is measured at 460 nm, I think.
Turbid samples, such as tissue homogenates, are not well-suited for colorimetric assays because of light scattering. If it doesn't work out, a radiometric assay (incorporation of radioactive acetate) would probably be your best bet.