## Model documentation 


**SMILES**: A single SMILES representing a drug for which the prediction should be performed.

**Multiple SMILES**: Alternatively, you can upload a `.smi` or a `.tsv` file that is tab-separated and contains SMILES in the first column. Note that it **must not** contain a header. Moreover, provide *either* a single SMILES *or* a file, not both!

**Transcriptomics data file**: Here, you can optionally upload an omics file with cell lines in rows and genes in columns. If not provided, predictions will be performed on the cell lines available in the [GDSC](https://academic.oup.com/nar/article/41/D1/D955/1059448) and [CCLE](https://sites.broadinstitute.org/ccle/) databases.

**Confidence**: This toggle determines whether the model returns confidence estimates. If toggled on, this will take ~15 times more time to run. The model will return two estimates, for aleatoric and epistemic uncertainty respectively.

## NOTE
If you are an user of the old, deprecated PaccMann webservice (that was hosted on IBM Cloud) and you miss certain functionalities such as analysing the SMILES or the gene attention, please reach out to {dow,jab,tte}@zurich.ibm.com and we will try to provide those features timely.


## Citation
If you use this webservice, please cite:
```bib
@article{cadow2020paccmann,
  title={PaccMann: a web service for interpretable anticancer compound sensitivity prediction},
  author={Cadow, Joris and Born, Jannis and Manica, Matteo and Oskooei, Ali and Rodr{\'\i}guez Mart{\'\i}nez, Mar{\'\i}a},
  journal={Nucleic acids research},
  volume={48},
  number={W1},
  pages={W502--W508},
  year={2020},
  publisher={Oxford University Press}
}
```