As can be seen, ComBat was successful in removing dataset-specific biases (Figure 2, Figure 3) since the samples of the merged data set integrated with ComBat were better intermixed than the samples merged with Z-score normalization.
We recomputed the accuracies for this removed dataset and compared them with those for the original dataset.
For this experiment the removed dataset was used to train the CRPhos model, whereas the remaining fraction was used for testing.
Yet paradoxically, data can be viewed as the opposite: an asset, an opportunity, co-created with walls removed, datasets connected, analysis undertaken, and lives improved.
We removed datasets that had only a few sequences after removing fragments with lengths less than 200 or more than 500 from the obtained FASTA files.
We first removed datasets that have less than 5 samples (e.g., tissues, conditions, or time points) to avoid possible spurious strong correlations.
Also, we removed datasets which have more than one percent of sequences including blacklist regions reported on https://sites.google.com/site/anshulkundaje/projects/blacklists.com/site/anshulkundaje/projects/blacklists
"However, as a result of the potential to extract some doctor and other service provider ID numbers, the Department of Health immediately removed the dataset from the website to ensure the security and integrity of the data is maintained".
DWD removes the dataset bias very efficiently while preserving the biological information.
§Seven participants removed from dataset for meta-analysis as they had preference for treatment not available in trial.
Once all CS duplicate sequences were removed, this dataset was merged with the remaining two sets, VV and WS.
