The thermal shift assay measures changes in fluorescence intensity of the dye as a function of temperature in the presence of the protein. If the ligand is also fluorescent, then the measured fluorescence will be the sum of the fluorescence intensities of the dye and ligand. As long as the ligand's fluorescent intensity is not too great compared with the dye's, it should be possible to subtract the thermogram of the ligand+protein from that of the ligand+dye+protein, assuming the same experimental procedure is used, everything is soluble, and there are no other artifacts.

A different problem would be if the ligand quenches or absorbs the fluorescence of the dye but is not itself fluorescent. In that case, you can measure the quenching effect and correct for it ratiometrically. For example, if you find that a certain ligand concentration reduces the dye's fluorescence by half, then the correction is to multiply the dye's fluorescence by 2. In the thermal shift assay, these correction factors could be determined by preparing a thermogram of the dye by itself and of the dye+ligand, both without the protein.

What do you mean by "I cannot find any gradient for IC50 determination?" Do you mean that varying the concentration of the ligand does not produce a correspondingly varying degree of inhibition? A colored compound can interfere in both absorbance an luminescence assays. Please see this paper for the method of correction.

Article Correction for Interference by Test Samples in High-Throughput Assays

There is an additional form of interference in the Malachite Green assay, which is interference with the Malachite Green dye binding to phosphate. When performing the artifact correction procedure, the concentration of phosphate used for the artifact assay should be set equal to one-half the amount produced in the enzyme reaction so that the correction at the IC50 concentration will be the most accurate.