Diagnostic biases in translational bioinformatics

BACKGROUND

With the surge of translational medicine and computational omics research, complex disease diagnosis is more and more relying on massive omics data-driven molecular signature detection. However, how to detect and prevent possible diagnostic biases in translational bioinformatics remains an unsolved problem despite its importance in the coming era of personalized medicine.

METHODS

In this study, we comprehensively investigate the diagnostic bias problem by analyzing benchmark gene array, protein array, RNA-Seq and miRNA-Seq data under the framework of support vector machines for different model selection methods. We further categorize the diagnostic biases into different types by conducting rigorous kernel matrix analysis and provide effective machine learning methods to conquer the diagnostic biases.

RESULTS

In this study, we comprehensively investigate the diagnostic bias problem by analyzing benchmark gene array, protein array, RNA-Seq and miRNA-Seq data under the framework of support vector machines. We have found that the diagnostic biases happen for data with different distributions and SVM with different kernels. Moreover, we identify total three types of diagnostic biases: overfitting bias, label skewness bias, and underfitting bias in SVM diagnostics, and present corresponding reasons through rigorous analysis. Compared with the overfitting and underfitting biases, the label skewness bias is more challenging to detect and conquer because it can be easily confused as a normal diagnostic case from its deceptive accuracy. To tackle this problem, we propose a derivative component analysis based support vector machines to conquer the label skewness bias by achieving the rivaling clinical diagnostic results.

CONCLUSIONS

Our studies demonstrate that the diagnostic biases are mainly caused by the three major factors, i.e. kernel selection, signal amplification mechanism in high-throughput profiling, and training data label distribution. Moreover, the proposed DCA-SVM diagnosis provides a generic solution for the label skewness bias overcome due to the powerful feature extraction capability from derivative component analysis. Our work identifies and solves an important but less addressed problem in translational research. It also has a positive impact on machine learning for adding new results to kernel-based learning for omics data.

Important options available--from start to finish--for translating proteomics results to clinical chemistry

In the realm of clinical chemistry, the field of clinical proteomics, that is, the application of proteomic methods for understanding mechanisms and enabling diagnosis, prediction, measurement of activity, and treatment response in disease, is first and foremost a discovery and research tool that feeds assay development downstream. Putative new assay candidates generated by proteomics discovery projects compete with well-established assays with known indications, well-described performance, and of known value in specific clinical settings. Careful attention to the many options available in the design, execution, and interpretation of clinical proteomics studies is thus necessary for translation into clinical practice. We here review and discuss important options associated with clinical proteomics endeavors stretching from the planning phases to the final use in clinical chemistry.

Overcome support vector machine diagnosis overfitting

Support vector machines (SVMs) are widely employed in molecular diagnosis of disease for their efficiency and robustness. However, there is no previous research to analyze their overfitting in high-dimensional omics data based disease diagnosis, which is essential to avoid deceptive diagnostic results and enhance clinical decision making. In this work, we comprehensively investigate this problem from both theoretical and practical standpoints to unveil the special characteristics of SVM overfitting. We found that disease diagnosis under an SVM classifier would inevitably encounter overfitting under a Gaussian kernel because of the large data variations generated from high-throughput profiling technologies. Furthermore, we propose a novel sparse-coding kernel approach to overcome SVM overfitting in disease diagnosis. Unlike traditional ad-hoc parametric tuning approaches, it not only robustly conquers the overfitting problem, but also achieves good diagnostic accuracy. To our knowledge, it is the first rigorous method proposed to overcome SVM overfitting. Finally, we propose a novel biomarker discovery algorithm: Gene-Switch-Marker (GSM) to capture meaningful biomarkers by taking advantage of SVM overfitting on single genes.

Disease Biomarker Query from RNA-Seq Data

As a revolutionary way to unveil transcription, RNA-Seq technologies are challenging bioinformatics for its large data volumes and complexities. A large number of computational models have been proposed for differential expression (DE) analysis and normalization from different standing points. However, there were no studies available yet to conduct disease biomarker discovery for this type of high-resolution digital gene expression data, which will actually be essential to explore its potential in clinical bioinformatics. Although there were many biomarker discovery algorithms available in traditional omics communities, they cannot be applied to RNA-Seq count data to seek biomarkers directly for its special characteristics. In this work, we have presented a biomarker discovery algorithm, SEQ-Marker for RNA-Seq data, which is built on a novel data-driven feature selection algorithm, nonnegative singular value approximation (NSVA), which contributes to the robustness and sensitivity of the following DE analysis by taking advantages of the built-in characteristics of RNA-Seq count data. As a biomarker discovery algorithm built on network marker topology, the proposed SEQ-Marker not only bridges transcriptomics and systems biology but also contributes to clinical diagnostics.