Integrating classifiers across datasets improves consistency of biomarker predictions for sepsis

Saraiva, J.; Oswald, Marcus; Biering, Antje; Assmann, Cora; Klassert, Tilman E.; Blaess, Markus; Czakai, Kristin; Claus, Ralf A.; Löffler, Jürgen; Slevogt, Hortense; König, Rainer

doi:10.1016/j.ifacol.2016.12.109

Cited by 2 publications

(11 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In our present study, no optimization on the number of features for classification was performed. No significant difference, in terms of performance and pairwise overlaps, was obtained when using 20 rather than 30 genes for classification in a previous study [ 15 ] which suggested that, while not optimal, a different selection of the constrained number of features would not lead to much different results.…”

Section: Discussionmentioning

confidence: 93%

“…Limiting the amount of time expended to obtain the optimized solution by use of solvers such as gurobi decreases the latter, while still presenting good solutions. To note, a previous study which we presented on a conference (see [ 15 ]) has demonstrated an initial implementation of our approach in the identification of consistent gene signatures capable of discriminating between infected and healthy samples. This was a more straight forward application when compared to the discrimination of the kind of infection (fungal versus bacterial).…”

Section: Discussionmentioning

confidence: 99%

“…For our study, we used data of a total of n = 27 samples which were extracted 3 and 6 h post-infection ( n = 9 bacterial and n = 18 fungal). We used transcriptomic data generated by Saraiva and co-workers [ 15 ] from healthy human isolated PBMCs which were challenged with either heat-killed C. albicans MYA-3573 yeast (MOI of 2) or LPS (10 ηg/ml) from E. coli 0111: B4 (InvivoGen). Four samples were extracted 4 h post-infection of each stimulus.…”

Section: Methodsmentioning

confidence: 99%

“…To tackle this issue, we employed a constrained based method based on Mixed Integer Linear Programming (MILP). MILPs have been used for the optimization of cell-network arrangements in order to discover patterns in pathways which are distinctively expressed [ 13 ], in the inference of gene regulation [ 14 ] and in the identification of gene signatures capable of distinguishing infected from non-infected samples [ 15 ]. A major advantage of applying our method resides on the reduction of the search space on which the optimization problem is performed by the imposition of constraints.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Fungal biomarker discovery by integration of classifiers

et al. 2017

Self Cite

View full text Add to dashboard Cite

BackgroundThe human immune system is responsible for protecting the host from infection. However, in immunocompromised individuals the risk of infection increases substantially with possible drastic consequences. In extreme, systemic infection can lead to sepsis which is responsible for innumerous deaths worldwide. Amongst its causes are infections by bacteria and fungi. To increase survival, it is mandatory to identify the type of infection rapidly. Discriminating between fungal and bacterial pathogens is key to determine if antifungals or antibiotics should be administered, respectively. For this, in situ experiments have been performed to determine regulation mechanisms of the human immune system to identify biomarkers. However, these studies led to heterogeneous results either due different laboratory settings, pathogen strains, cell types and tissues, as well as the time of sample extraction, to name a few.MethodsTo generate a gene signature capable of discriminating between fungal and bacterial infected samples, we employed Mixed Integer Linear Programming (MILP) based classifiers on several datasets comprised of the above mentioned pathogens.ResultsWhen combining the classifiers by a joint optimization we could increase the consistency of the biomarker gene list independently of the experimental setup. An increase in pairwise overlap (the number of genes that overlap in each cross-validation) of 43% was obtained by this approach when compared to that of single classifiers. The refined gene list was composed of 19 genes and ranked according to consistency in expression (up- or down-regulated) and most of them were linked either directly or indirectly to the ERK-MAPK signalling pathway, which has been shown to play a key role in the immune response to infection. Testing of the identified 12 genes on an unseen dataset yielded an average accuracy of 83%.ConclusionsIn conclusion, our method allowed the combination of independent classifiers and increased consistency and reliability of the generated gene signatures.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-017-4006-x) contains supplementary material, which is available to authorized users.

show abstract

Section: Discussionmentioning

confidence: 93%

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Fungal biomarker discovery by integration of classifiers

et al. 2017

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the generated gene signatures from independent studies usually do not present a high degree of consistency even if the same discrimination problem was addressed. We previously showed that combining classifiers using a Mixed Integer Linear Programming (MILP) improved consistency of gene signatures even if generated from quite diverse settings (Saraiva et al, 2016 ). The gene signature produced by Saraiva et al accurately discriminated infected from non-infected samples with an average accuracy of 92% and was proposed as a generic host immune response toward infections due to the heterogeneity of the expression datasets in terms of immune cell stimulation.…”

Section: Introductionmentioning

confidence: 99%

Combination of Classifiers Identifies Fungal-Specific Activation of Lysosome Genes in Human Monocytes

et al. 2017

Self Cite

View full text Add to dashboard Cite

Blood stream infections can be caused by several pathogens such as viruses, fungi and bacteria and can cause severe clinical complications including sepsis. Delivery of appropriate and quick treatment is mandatory. However, it requires a rapid identification of the invading pathogen. The current gold standard for pathogen identification relies on blood cultures and these methods require a long time to gain the needed diagnosis. The use of in situ experiments attempts to identify pathogen specific immune responses but these often lead to heterogeneous biomarkers due to the high variability in methods and materials used. Using gene expression profiles for machine learning is a developing approach to discriminate between types of infection, but also shows a high degree of inconsistency. To produce consistent gene signatures, capable of discriminating fungal from bacterial infection, we have employed Support Vector Machines (SVMs) based on Mixed Integer Linear Programming (MILP). Combining classifiers by joint optimization constraining them to the same set of discriminating features increased the consistency of our biomarker list independently of leukocyte-type or experimental setup. Our gene signature showed an enrichment of genes of the lysosome pathway which was not uncovered by the use of independent classifiers. Moreover, our results suggest that the lysosome genes are specifically induced in monocytes. Real time qPCR of the identified lysosome-related genes confirmed the distinct gene expression increase in monocytes during fungal infections. Concluding, our combined classifier approach presented increased consistency and was able to “unmask” signaling pathways of less-present immune cells in the used datasets.

show abstract

Integrating classifiers across datasets improves consistency of biomarker predictions for sepsis

Cited by 2 publications

References 32 publications

Fungal biomarker discovery by integration of classifiers

Fungal biomarker discovery by integration of classifiers

Combination of Classifiers Identifies Fungal-Specific Activation of Lysosome Genes in Human Monocytes

Contact Info

Product

Resources

About