# DDA-bench: a manually curated database for benchmarking datasets and baseline performance values in predicting drug-disease associations

## TL;DR

DDA-Bench is a curated database that provides benchmark datasets and baseline performance values for predicting drug-disease associations, saving time for researchers.

## Contribution

The novel contribution is the creation of a manually curated database with benchmark datasets and baseline performance values for drug-disease association prediction.

## Key findings

- Performance variations for a given method across different reports should be considered.
- Dataset density impacts predictive performance and should be addressed in future studies.
- DDA-Bench is publicly available at https://dda.csbios.net.

## Abstract

Predicting drug-diseases associations provides hints in developing new drugs. Various computational methods have been developed. To develop better models for predicting drug-disease associations, two types of key resources must be obtained, the benchmarking dataset and the baseline performances. Collecting these resources usually requires extensive labors in reading literatures and extracting relevant information from the literature manually. We developed DDA-Bench databases services, which curates commonly used benchmarking datasets and up-to-date performance values from baseline studies. We analyzed data records in DDA-Bench database. We proposed that performance variations for a given method in the context of different reports should be noticed. The impact of dataset density on predictive performance exists, and should be considered in future studies. In addition, we release the DDA-Bench database to the public. The DDA-Bench database saves time and efforts in constructing data basis for developing new models for predicting drug-disease associations. The DDA-Bench database can be accessed at https://dda.csbios.net.