How to participate to the challenge

1. Download data

You are offered both training-set and known test data-set to evaluate quality of your models. There is also a blind test-set that will be used to determine a winner. The experimental values for the blind dataset will be available to the Organising Committee and Advisory Board after September 1st. The experimental toxicities for these molecules have not been yet published. The data-sets are provided in several formats. To download the data and get further information, go here.

2. Develop models

Proposed work-flow for model development

- Develop model using the training set
- Predict molecules from the known test set
- Predict molecules from the blind data set using the previous model or new model developed using all available data (i.e., training and known data set)

3. Upload and register

Prepare a file with predicted values and (optionally) their predicted errors for each molecule from known and blind test sets. A detailed description what the file should look like can be found here.

Upload the file and register your team . You will receive a password (to be sent to your e-mail address provided during registration).

4. Check your results

After few seconds a ranking of your approach according to its performance for the known test set will be provided.

You can upload new predictions as many times as you want. The results of the last prediction only will be used for the final evaluation. The ranking for the known test set is a preliminary one and will not be used for the final scoring. The results for the blind set and winners of the challenge will be announced in September.