A deep learning approach for predicting the active phytochemical constituent’s presence in Premna Latifolia Leaf Extracts Data set.

Abstract

In this work the Premna latifolia (P.Latifolia) leaf extracts active phytochemical constituents presence were predicted using deep learning method. The computation techniques with bioinformatics data enables knowledge extractions from different sources. These interdisciplinary focus would reduce human efforts, efficiently increase the accuracy of the prediction results. In this work, the bioinformatics data transformed to valuable knowledge using machine learning model. The active phytochemical constituent’s presence in P.Latifolia leaf extracts will be helpful in preparation of herbal medicine for antibacterial, antioxidant, wound healing and anti-inflammatory activities. P.Latifolia leaf extracts has shown the potential inhibitory actions, with several solvents such as hexane, ethyl acetate, methanol, aqueous etc. In our work, preparation of leaf extracts at the concentration of 100µg/ml was taken for further studies. Identified phytochemical constituents are Alkaloids, Glycosides, Saponins, Phenolic-compounds, Tannins, Proteins, Flavonoids, Terpenoids are considered for this study but the list is so many. The label attribute ‘presence’, which classifies the result as Very High, High, Medium, Low and Very Low as based on the average number of phyto constituents present in each solvents These phytochemical constituents data are binomial with 245 tuples with 10 attributes are given as input to the deep learning model, designed using rapid miner tool and observed the result for five runs. The Accuracy of the model was observed at the average of accuracy: 97.52% +/- 4.47% (micro average: 97.55%) in three runs after 12 seconds gap for each run.