Data Set Information:

药物通常是小的有机分子，通过与受体上的靶点结合而达到预期的活性。发现新药的第一步通常是识别和分离与其结合的受体，然后测试许多小分子与靶点结合的能力。这使得研究人员的任务是确定活性（结合）化合物和非活性（非结合）化合物的区别。这样的测定可用于设计新的化合物，这些化合物不仅结合，而且还具有药物所需的所有其他性质（溶解度、口服吸收、无副作用、适当的作用时间、毒性等）。

为了进行特征选择挑战，修改了原始数据。特别是，我们增加了一些分心功能，称为“探针”，没有预测能力。特征和模式的顺序是随机的。

DOROTHEA -- Positive ex. -- Negative ex. -- Total  
Training set -- 78 -- 722 -- 800  
Validation set -- 34 -- 316 -- 350  
Test set -- 78 -- 722 -- 800  
All -- 190 -- 1760 -- 1950

We mapped Active compounds to the target value +1 (positive examples) and Inactive compounds to the target value –1 (negative examples).

Number of variables/features/attributes:
Real: 50000
Probes: 50000
Total: 100000

This dataset is one of five datasets used in the NIPS 2003 feature selection challenge. Our website [Web link] is still open for post-challenge submissions. Information about other related challenges are found at: [Web link]. The CLOP package includes sample code to process these data: [Web link].

All details about the preparation of the data are found in our technical report: Design of experiments for the NIPS 2003 variable selection benchmark, Isabelle Guyon, July 2003, [Web link] (also included in the dataset archive). Such information was made available only after the end of the challenge.

The data are split into training, validation, and test set. Target values are provided only for the 2 first sets. Test set performance results are obtained by submitting prediction results to: [Web link].

The data are in the following format:
dataname.param: Parameters and statistics about the data
dataname.feat: Identities of the features (withheld, to avoid biasing feature selection).
dataname_train.data: Training set (a sparse binary matrix, patterns in lines, features in columns: the number of the non-zero features are provided).  
dataname_valid.data: Validation set.    
dataname_test.data: Test set.
dataname_train.labels: Labels (truth values of the classes) for training examples.  
dataname_valid.labels: Validation set labels (withheld during the benchmark, but provided now).
dataname_test.labels: Test set labels  (withheld, so the data can still be use as a benchmark).

Attribute Information:

We do not provide attribute information to avoid biasing feature selection.

Relevant Papers:

The best challenge entrants wrote papers collected in the book:
Isabelle Guyon, Steve Gunn, Masoud Nikravesh, Lofti Zadeh (Eds.), Feature Extraction, Foundations and Applications. Studies in Fuzziness and Soft Computing. Physica-Verlag, Springer. [Web link]

See also:
Isabelle Guyon, et al, 2007. Competitive baseline methods set new standards for the NIPS 2003 feature selection benchmark. Pattern Recognition Letters 28 (2007) 1438–1444.
and the associated technical report:
Isabelle Guyon, et al. 2006. Feature selection with the CLOP package. Technical Report. [Web link].

Citation Request:

Isabelle Guyon, Steve R. Gunn, Asa Ben-Hur, Gideon Dror, 2004. Result analysis of the NIPS 2003 feature selection challenge. In: NIPS. [Web link].

a. Original owners
The dataset with which DOROTHEA was created is one of the KDD (Knowledge Discovery in Data Mining) Cup 2001. The original dataset and papers of the winners of the competition are available at: http://www.cs.wisc.edu/~dpage/kddcup2001/. DuPont Pharmaceuticals graciously provided this data set for the KDD Cup 2001 competition. All publications referring to analysis of this data set should acknowledge DuPont Pharmaceuticals Research Laboratories and KDD Cup 2001.

b. Donor of database
This version of the database was prepared for the NIPS 2003 variable and feature selection benchmark by Isabelle Guyon, 955 Creston Road, Berkeley, CA 94708, USA (isabelle '@' clopinet.com).