1 results listed
Publicly available compound and bioactivity
databases, such as ChEMBL, provide the main foundation for
data-driven drug screening. In this work, we were interested
in the Prostate-Specific Antigen (PSA), which is the most
important biomarker in the context of the diagnostics and the
therapy of prostate cancer. The dataset comprised about 95%
of high-confidence bioactivity data, it is the major resource
for construction of a good Random Forest classification
model. After connecting PSA-targeted bioactivity data with
the molecular descriptors and genomic counts, a machine
learning model was developed, that was then able to predict
bioactive compounds with high accuracy. After cleaning,
extracting the features from the data, together with training of
the machine learning model we have got the most reliable
prediction. Our Data reveals that the introduction of genomic
and structural features notably enhances prediction
performance using the traditional QSAR approach. The final
mode continuously can not only identify active, but also
inactive compounds with high confidence and reliability.
This method not only simplifies early drug discovery but also
shows the strength of AI-driven methods that can identify
those drugs that are expected to be the most effective PSA
inhibitors for prostate cancer therapy.
International Conference on Advanced Technologies, Computer Engineering and Science
ICATCES
Manjunath Managuli
Swetha Goudar
Sangmesh C Managuli