Cg is the most affected one, because it depends on the amount of charges in graphene. Supposing the strain introduces a bandgap, you will have less charge carriers in equilibrium state, meaning a smaller quantum capacitance Cq, thus a smaller total Cg.

Source and drain capacitors instead can be considered only dependent on the geometry, but this in general depends on the type of model you are using. Ballistic models can use Cs and Cd as a simplified way to compute the uncharged electrostatic potential of the channel.

Thanks Giancarlo

I appreciate you if I have some numeric amount of Cg and other capacitor?

Upon applying strain, the work function of the graphene ribbon in the channel will change, and this will affect the the metal-GNR contact. For strain effect on work function of GNR, there seems to be no report yet. However, i have calculated the evolution of work function of CNTs, this may give you some hints. (Strain effects on work functions of pristine and potassium-decorated, JCP, 2009)