Limits...
Lung necrosis and neutrophils reflect common pathways of susceptibility to Mycobacterium tuberculosis in genetically diverse, immune-competent mice.

Niazi MK, Dhulekar N, Schmidt D, Major S, Cooper R, Abeijon C, Gatti DM, Kramnik I, Yener B, Gurcan M, Beamer G - Dis Model Mech (2015)

Bottom Line: Low to no IFN-γ, IL-12, IL-2 and IL-10 successfully discriminated non-infected mice from infected mice but failed to discriminate disease status amongst supersusceptible, susceptible and resistant M.-tuberculosis-infected DO mice.Additional analyses identified CXCL1 as a promising peripheral biomarker of disease and of CXCL1 production in the lungs.From these results, we conclude that: (1) DO mice respond variably to M. tuberculosis infection and will be useful to identify pathways involving necrosis and neutrophils; (2) data from DO mice is suited for machine learning methods to build, validate and test models with independent data based solely on molecular biomarkers; (3) low levels of immunological cytokines best indicate a lack of exposure to M. tuberculosis but cannot distinguish infection from disease.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Informatics, The Ohio State University, Columbus, 43210 OH, USA.

No MeSH data available.


Related in: MedlinePlus

Classification tree based on neutrophil chemokines CXCL1, CXCL2, CXCL5. Female 8-week-old non-sibling DO mice and C57BL/6J mice were infected with ∼100 M. tuberculosis bacilli by aerosol. From the first experiment, complete data for all 15 molecular parameters used for model generation and training were obtained from 70 M.-tuberculosis-infected DO mice and eight non-infected DO mice. The best-performing classification tree used only lung and plasma CXCL1, CXCL2, CXCL5 (A), which was then validated by the leave-one-out method (B). After validation, performance of the classification tree was stringently tested using data from the second independent experiment, which included 52 M.-tuberculosis-infected DO mice, three non-infected DO mice and five M.-tuberculosis-infected C57BL/6J mice (C), with complete data from all 15 molecular parameters.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4582107&req=5

DMM020867F4: Classification tree based on neutrophil chemokines CXCL1, CXCL2, CXCL5. Female 8-week-old non-sibling DO mice and C57BL/6J mice were infected with ∼100 M. tuberculosis bacilli by aerosol. From the first experiment, complete data for all 15 molecular parameters used for model generation and training were obtained from 70 M.-tuberculosis-infected DO mice and eight non-infected DO mice. The best-performing classification tree used only lung and plasma CXCL1, CXCL2, CXCL5 (A), which was then validated by the leave-one-out method (B). After validation, performance of the classification tree was stringently tested using data from the second independent experiment, which included 52 M.-tuberculosis-infected DO mice, three non-infected DO mice and five M.-tuberculosis-infected C57BL/6J mice (C), with complete data from all 15 molecular parameters.

Mentions: By using machine learning methods, we extracted the same molecular features that were strong disease correlates or had some statistical ability to distinguish classes: lung CXCL1, CXCL2, CXCL5, TNF, IFN-γ, IL-12; and two blood features – IL-2 and TNF. Although our data also contained three disease indicators (survival, percentage of peak body weight at euthanasia and M. tuberculosis CFU), these were purposefully excluded because the goal was to identify molecules capable of classifying disease status and discrimination of a non-infected state. For building and testing classification models, only data from individual mice with complete data – i.e. all 15 molecular features – were used. Thus, the training and model validation steps used N=70 DO mice from the first experiment that had been infected with 127±68 M. tuberculosis Erdman bacilli and N=8 non-infected DO mice. Multiple machine learning approaches were used, listed in the Materials and Methods. Of these, classification trees were pursued because they produced the most accurate models. Validation for the model used the standard leave-one-out cross-validation for training data (Fig. 4B). However, performance rates in training cannot be generalized. A more stringent approach to evaluate performance (accuracy in classifying disease status of DO and C57BL/6J mice) is to test the model using completely independent molecular data. Therefore, we tested the models using data from a completely independent experiment, again using only mice for which all 15 molecular parameters were available. This included N=55 DO mice and N=5 C57BL/6J mice from the second experiment that had been infected with 97±61 bacilli and N=3 non-infected DO mice.Fig. 4.


Lung necrosis and neutrophils reflect common pathways of susceptibility to Mycobacterium tuberculosis in genetically diverse, immune-competent mice.

Niazi MK, Dhulekar N, Schmidt D, Major S, Cooper R, Abeijon C, Gatti DM, Kramnik I, Yener B, Gurcan M, Beamer G - Dis Model Mech (2015)

Classification tree based on neutrophil chemokines CXCL1, CXCL2, CXCL5. Female 8-week-old non-sibling DO mice and C57BL/6J mice were infected with ∼100 M. tuberculosis bacilli by aerosol. From the first experiment, complete data for all 15 molecular parameters used for model generation and training were obtained from 70 M.-tuberculosis-infected DO mice and eight non-infected DO mice. The best-performing classification tree used only lung and plasma CXCL1, CXCL2, CXCL5 (A), which was then validated by the leave-one-out method (B). After validation, performance of the classification tree was stringently tested using data from the second independent experiment, which included 52 M.-tuberculosis-infected DO mice, three non-infected DO mice and five M.-tuberculosis-infected C57BL/6J mice (C), with complete data from all 15 molecular parameters.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4582107&req=5

DMM020867F4: Classification tree based on neutrophil chemokines CXCL1, CXCL2, CXCL5. Female 8-week-old non-sibling DO mice and C57BL/6J mice were infected with ∼100 M. tuberculosis bacilli by aerosol. From the first experiment, complete data for all 15 molecular parameters used for model generation and training were obtained from 70 M.-tuberculosis-infected DO mice and eight non-infected DO mice. The best-performing classification tree used only lung and plasma CXCL1, CXCL2, CXCL5 (A), which was then validated by the leave-one-out method (B). After validation, performance of the classification tree was stringently tested using data from the second independent experiment, which included 52 M.-tuberculosis-infected DO mice, three non-infected DO mice and five M.-tuberculosis-infected C57BL/6J mice (C), with complete data from all 15 molecular parameters.
Mentions: By using machine learning methods, we extracted the same molecular features that were strong disease correlates or had some statistical ability to distinguish classes: lung CXCL1, CXCL2, CXCL5, TNF, IFN-γ, IL-12; and two blood features – IL-2 and TNF. Although our data also contained three disease indicators (survival, percentage of peak body weight at euthanasia and M. tuberculosis CFU), these were purposefully excluded because the goal was to identify molecules capable of classifying disease status and discrimination of a non-infected state. For building and testing classification models, only data from individual mice with complete data – i.e. all 15 molecular features – were used. Thus, the training and model validation steps used N=70 DO mice from the first experiment that had been infected with 127±68 M. tuberculosis Erdman bacilli and N=8 non-infected DO mice. Multiple machine learning approaches were used, listed in the Materials and Methods. Of these, classification trees were pursued because they produced the most accurate models. Validation for the model used the standard leave-one-out cross-validation for training data (Fig. 4B). However, performance rates in training cannot be generalized. A more stringent approach to evaluate performance (accuracy in classifying disease status of DO and C57BL/6J mice) is to test the model using completely independent molecular data. Therefore, we tested the models using data from a completely independent experiment, again using only mice for which all 15 molecular parameters were available. This included N=55 DO mice and N=5 C57BL/6J mice from the second experiment that had been infected with 97±61 bacilli and N=3 non-infected DO mice.Fig. 4.

Bottom Line: Low to no IFN-γ, IL-12, IL-2 and IL-10 successfully discriminated non-infected mice from infected mice but failed to discriminate disease status amongst supersusceptible, susceptible and resistant M.-tuberculosis-infected DO mice.Additional analyses identified CXCL1 as a promising peripheral biomarker of disease and of CXCL1 production in the lungs.From these results, we conclude that: (1) DO mice respond variably to M. tuberculosis infection and will be useful to identify pathways involving necrosis and neutrophils; (2) data from DO mice is suited for machine learning methods to build, validate and test models with independent data based solely on molecular biomarkers; (3) low levels of immunological cytokines best indicate a lack of exposure to M. tuberculosis but cannot distinguish infection from disease.

View Article: PubMed Central - PubMed

Affiliation: Department of Biomedical Informatics, The Ohio State University, Columbus, 43210 OH, USA.

No MeSH data available.


Related in: MedlinePlus