A special class of
cell null boundary invertible three neighborhood CA referred to as Equal Length Cycle CA (ELCCA) is proposed in this paper to represent the features of
bit symbol strings. Necessary and sufficient conditions for generation of ELCCA has been reported. A specific set of ELCCA cycles are selected by employing the mRMR algorithm  popularly used for feature extraction of symbol strings. An algorithm is next developed to classify the symbol strings based on the feature set extracted. The proposed CA model has been validated for analyzing symbol string of biomolecules referred to as Enzymes. These biomolecules are classified on the basis of the catalytic reaction they participate. The symbol string classification algorithm predicts the class of any input enzyme with accuracy varying from 90.4% to 98.6%. Experimental results have been reported for 22800 enzymes with wide variation in species.