Mousumi Basu, A typical curve shape here is a sigmoid. To describe it, one can use e.g. compositon of exponential and power functions or composition of hyperbolic and power functions. I would avoid transforming the nonlinear functions so that in the new coordinate system the experimental points are arranged along a straight line. Better to use non=linear methods commonly called non-linear regression. And one more thing - such a transformation is not linearization.