Using a morphological imaging dataset from Recursion Pharmaceuticals, this exploratory work explores the use of deep learning in pharmaceutical drug development. The dataset, which was created in April 2023, includes 309,522 5-channel images showing how cells were treated in 1738 microplates. Three different cell treatments are included in the research methodology: fake cells, cells containing irradiated SARS-CoV-2 virons, and cells infected with active SARS-CoV-2 that have been treated with drug library candidates. The resulting RxRx19a dataset is an essential tool for studying the effects of compounds on human kidney cells infected with SARS-CoV-2. It is organised by FDA, EMA, and clinical trial chemicals. The cytopathic character of the SARS-CoV-2 virus is shown by electron microscopy data, which advise the utilization regarding persistent renal trade treatment for Coronavirus patients who have intense kidney injury. Additionally, the Rx1 dataset—which is treated with siRNA rather than compounds—is included in the article, facilitating the first training of a classification model. The SARS-CoV-2 dataset's specific viral circumstances are used to fine-tune the model training strategy, which is a two-step process that uses the heterogeneous siRNA dataset. Experiments show that DenseNet is better than other deep neural network designs at classifying siRNA pictures, and on the SARS-CoV-2 dataset, a cascade transfer learning model performs better than other models. The model is then used to rank the effectiveness of various drugs in treating COVID-19, thereby finding possible research prospects. This thorough investigation highlights the potential of deep learning to improve drug discovery, especially when it comes to viral illnesses such as SARS-CoV-2.