kexinhuang12345 / DeepPurpose

A Deep Learning Toolkit for DTI, Drug Property, PPI, DDI, Protein Function Prediction (Bioinformatics)
https://doi.org/10.1093/bioinformatics/btaa1005
BSD 3-Clause "New" or "Revised" License
962 stars 272 forks source link

Broad Repurposing Drug Name question #103

Closed BAREJAA closed 3 years ago

BAREJAA commented 3 years ago

What do the values in the "Drug Name" column of the table produced by models.repurpose refer to? For example, my top hit is "5063.0", but I can't tell what drug this value refers to. I don't see it anywhere in the data table I downloaded from the Drug Repurposing Hub.

Thanks!

kexinhuang12345 commented 3 years ago

Hey, sorry for the confusion, the Drug Name column is created for data that have IDs. But for many, when no IDs are found in the dataset, we simply generate a drug index by the "X_repurpose". You can also feed your drug name list by specifying "drug_names", a list of names matching with X_repurpose.

BAREJAA commented 3 years ago

So, in the table below, does "Drug Name" refer to the CID?

Drug Repurposing Result for SDHA +------+-------------+-------------+---------------+ | Rank | Drug Name | Target Name | Binding Score | +------+-------------+-------------+---------------+ | 1 | 11434515.0 | SDHA | 8.34 | | 2 | 11598628.0 | SDHA | 7.79 | | 3 | 11713159.0 | SDHA | 7.71 |

BAREJAA commented 3 years ago

Another question (sorry for all the questions!) - since I used the 'CNN_CNN_BindingDB_IC50' pre-trained model, does that mean the "Binding Score" column in the table above contains pIC50 values?

kexinhuang12345 commented 3 years ago

yes, this looks like CIDs. You can specify the model to output pIC50 or IC50 by 'convert_y = True' in default, it is set to be recovered back to the IC50 but the model is trained on pIC50

BAREJAA commented 3 years ago

Thanks again!