The technology is evolving quickly, and the immense and complex protein
sequence databases are growing every day. To extract information from these huge
masses of data, we require sophisticated computational approaches, such as Quantum
Machine Learning (QML), which could be rapidly applied for the benefit of humanity.
The HPFP (Human-Protein-Function-Prediction) is a crucial study used for the
discovery of disease diagnosis, medicines, and crop hybridization. These days, lots of
computational approaches are available for HPFP, yet many do not deliver reliable
findings rapidly. This research study performed the experiment for HPFP by extracting
the dataset from the HPRD (Human-Protein-Reference-Database), the SDFs
(Sequence-Derived-Features) from protein sequence are extracted through the proposed
HP-SDFES, as well as from some other web-servers. This study utilizes an RF
(Random Forest) Machine Learning (ML) classifier for HPFP by extracting SDFs from
HPF (Human-Protein-Sequence) and achieved 74.91% accuracy for individual protein
class prediction. With improved SDF extractions and in conjunction with input
configurations, the prediction accuracy enhances and ranges from 98% to 100% under
ideal conditions. This study identifies that RF is the better classifier for HPFP rather
than any other ML classifier, and this methodology can be applied similarly to other
allied studies for the advancement of humanity, which is also the key finding of this
research study. Further, to increase the prediction process speed, the QML approach
has been used on RF. Quantum computing improves RF by accelerating feature selection, entropy computations, and decision tree splits. The obtained final result has
been compared with the existing benchmarks.
Keywords: Decision tree, Human protein function (HPF), HPRD, Hybrid quantum-classical models, Mutation prediction, Prediction accuracy, Protein function prediction (PFP), Protein-protein interaction (PPI), Random forest classifier, Quantum feature map, Quantum kernel, Quantum machine learning (QML), SDF, ZZFeatureMap.
