I have been trying to identify correlations between assay specific EC50 values to in silico data (binding free energy, residue specific hydrogen bond occupancy, MM/PBSA energy values) for a range of compounds and their target receptors. Since they are different parametric dataset, I am looking forward to a more non-linear correlation approach. However, state-of-the-art approaches like KendallTau, Cannonical Correlation is unable to explain any variability or likelihood among the in silico and in vitro data. Dimensionality reduction method also does not explain much variability. This could be because I am trying to compare two different domains of data.
I would like to know if there is any established method of comparing such datasets.
You can plot EC50 vs in silico data in x and y axis
Try to see any correlation between them
You may perform a comparative analysis between the datasets. This can involve identifying trends, patterns, or similarities in the data that may provide insights into the relationship between the in silico predictions and the assay-specific EC50 values. Ensure that you have appropriate validation strategies in place to assess the robustness and generalizability of any correlations or models that you develop. Cross-validation techniques can help evaluate the performance of the models and provide an indication of their reliability.
Quite a difficult question to reply. You can find a tentative reply, possibly applicable only to the specific compounds investigated and not necessarily exportable to different compounds, in this paper:
Article Structure-Activity Relationships in Cytotoxic Au-I/Au-III Co...
- Badges
- Science topic