Despite a large number of in-situ missions, only VeGa-2 provided the high-resolution thermal measurements of the deep atmosphere of Venus below 12 km and they indicated a highly unstable layer below 7 km altitude. Lebonnois and Schubert (2017) tried to explain the VeGa-2 measurements by introducing a theory of a composition gradient in the deep atmosphere of Venus. In our research, we test this theory of composition gradient by making use of the near-infrared observations from Akatsuki and Venus Express missions in combination with the altimetry observations from Magellan Mission. To make use of the highly contaminated Akatsuki dataset, we develop a procedure to sequentially reduce the contamination to an acceptable level. Next, an atmospheric radiative transfer model is built to simulate the thermal emission coming from the surface of Venus, which is then used to generate a temperature vs altitude profile from the observations. Using this temperature profile we investigate the theory of composition gradient.