Estimating Bioinformatics Prediction Reliability
Author Information
Author(s): Oliviero Carugo
Primary Institution: Pavia University and Max F. Perutz Laboratories, Vienna University
Hypothesis
The reliability of bioinformatics predictions depends on the amount of data used for predictions.
Conclusion
Using Fragmented Prediction Performance Plots (FPPPs) is essential to quantify the performance of bioinformatics prediction protocols.
Supporting Evidence
- The reliability of predictions improves with larger datasets.
- FPPPs can help determine if a prediction method has reached its best performance.
- Different prediction methods may behave differently based on the amount of data.
Takeaway
This study shows that the more data you have, the better your predictions can be, but you need to check if adding more data will actually help.
Methodology
The study describes a tool called Fragmented Prediction Performance Plot (FPPP) to analyze the relationship between prediction reliability and data amount.
Potential Biases
The predictions may be influenced by the specific characteristics of the training sets used.
Limitations
The analysis may not account for all factors affecting prediction quality, such as data noise and heterogeneity.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website