Abstract
Adaptor proteins, pivotal in signal transduction ,consist of diverse modular domains, each exhibiting unique binding activities, forming complexes with intracellular signaling molecules. Implications of adaptor proteins in various human diseases underscore the need for accurate predictive models. In addressing this, we compiled a dataset featuring 2,484 positive (G0:0060090) and 15,495 negative (G0:0140110) results. Removal of highly similar sequences using the bio-conda CDHIT API yielded 1429 non-redundant clustered Adaptor proteins for G0:0060090. Similarly, G0:0140110 resulted in 8076 non-redundant clustered Adaptor proteins. Employing a 5-step rule predictor based on statistical moments and PseAAC for feature extraction, we split the dataset into 80% training and 20% testing. Our approach, currently employing known neutral models, advances bioinformatics efforts in anticipating the actions of adaptor proteins, holding promise for unravelingintricate cellular signaling mechanisms.