Summary
This update rectifies a numerical error regarding the dataset size used in the SVSBI study. The authors clarified that the protein-protein interaction (PPI) dataset consisted of 1,695 complexes, rather than the 1,795 originally reported, and have updated the supplementary files accordingly.
Context and Problem
The original research, SVSBI (Sequence-Based Virtual Screening of Biomolecular Interactions), addresses the challenge of predicting molecular interactions using computational methods. By focusing on sequence-based screening, the method provides a computationally efficient alternative to techniques requiring complex 3D structural data.
This correction addresses a discrepancy in data reporting. In computational biology, the exact size and composition of a training or validation dataset are critical. Even a small variation in sample size—in this case, a difference of 100 samples (approximately 5.5%)—can impact the perceived performance of a model or the ability of other researchers to replicate results. By issuing this correction, the authors ensure that the community has the precise parameters needed to evaluate or build upon their work.
Data Curation and Methodology
The primary dataset mentioned in this correction is derived from the PDBbind database, a well-known repository of experimentally measured binding affinities for biomolecular complexes. The correction clarifies the curation process as follows:
- Corrected Count: 1,695 PPI complexes.
- Original (Erroneous) Count: 1,795 PPI complexes.
The authors specified that these 1,695 samples were selected based on having "only two different sub-chain sequences," a specific filter used to ensure the quality and consistency of the data used for the virtual screening model. The correction confirms that the revised version of the supplementary file has replaced the original to reflect these changes.
Key Insights
- Data Integrity is Paramount: Accurate reporting of sample size ($N$) is the foundation of statistical significance and scientific rigor. This correction serves as a reminder of the rigors of peer-reviewed publishing.
- Transparency in Curation: The correction highlights the specific criteria used for the dataset (e.g., the requirement for two different sub-chain sequences), which is essential for understanding the model's scope and limitations.
- Supplementary Material Updates: The correction notes that the supplementary files—often containing the raw data or code used to generate results—have been synchronized with the corrected text.
Conclusion
The authors state that the original article has been corrected to reflect these numbers, implying that the core conclusions and the performance of the SVSBI model remain valid despite the correction of these specific figures. The update ensures the integrity and reproducibility of the computational models in the bioinformatics record.
Summary updated