Statistical prediction of immunity to placental malaria based on multi-assay antibody data for malarial antigens
Research output: Contribution to journal › Journal article › Research › peer-review
Standard
Statistical prediction of immunity to placental malaria based on multi-assay antibody data for malarial antigens. / Siriwardhana, Chathura; Fang, Rui; Salanti, Ali; Leke, Rose G. F.; Bobbili, Naveen; Taylor, Diane Wallace; Chen, John J.
In: Malaria Journal, Vol. 16, 391, 2017.Research output: Contribution to journal › Journal article › Research › peer-review
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - JOUR
T1 - Statistical prediction of immunity to placental malaria based on multi-assay antibody data for malarial antigens
AU - Siriwardhana, Chathura
AU - Fang, Rui
AU - Salanti, Ali
AU - Leke, Rose G. F.
AU - Bobbili, Naveen
AU - Taylor, Diane Wallace
AU - Chen, John J.
PY - 2017
Y1 - 2017
N2 - BackgroundPlasmodium falciparum infections are especially severe in pregnant women because infected erythrocytes (IE) express VAR2CSA, a ligand that binds to placental trophoblasts, causing IE to accumulate in the placenta. Resulting inflammation and pathology increases a woman’s risk of anemia, miscarriage, premature deliveries, and having low birthweight (LBW) babies. Antibodies (Ab) to VAR2CSA reduce placental parasitaemia and improve pregnancy outcomes. Currently, no single assay is able to predict if a woman has adequate immunity to prevent placental malaria (PM). This study measured Ab levels to 28 malarial antigens and used the data to develop statistical models for predicting if a woman has sufficient immunity to prevent PM.MethodsArchival plasma samples from 1377 women were screened in a bead-based multiplex assay for Ab to 17 VAR2CSA-associated antigens (full length VAR2CSA (FV2), DBL 1-6 of the FCR3, 3D7 and 7G8 lines, ID1-ID2a (FCR3 and 3D7) and 11 antigens that have been reported to be associated with immunity to P. falciparum (AMA-1, CSP, EBA-175, LSA1, MSP1, MSP2, MSP3, MSP11, Pf41, Pf70 and RESA)). Ab levels along with clinical variables (age, gravidity) were used in the following seven statistical approaches: logistic regression full model, logistic regression reduced model, recursive partitioning, random forests, linear discriminant analysis, quadratic discriminant analysis, and support vector machine.ResultsThe best and simplest model proved to be the logistic regression reduced model. AMA-1, MSP2, EBA-175, Pf41, and MSP11 were found to be the top five most important predictors for the PM status based on overall prediction performance.ConclusionsNot surprising, significant differences were observed between PM positive (PM+) and PM negative (PM−) groups for Ab levels to the majority of malaria antigens. Individually though, these malarial antigens did not achieve reasonably high performances in terms of predicting the PM status. Utilizing multiple antigens in predictive models considerably improved discrimination power compared to individual assays. Among seven different classifiers considered, the reduced logistic regression model produces the best overall predictive performance.
AB - BackgroundPlasmodium falciparum infections are especially severe in pregnant women because infected erythrocytes (IE) express VAR2CSA, a ligand that binds to placental trophoblasts, causing IE to accumulate in the placenta. Resulting inflammation and pathology increases a woman’s risk of anemia, miscarriage, premature deliveries, and having low birthweight (LBW) babies. Antibodies (Ab) to VAR2CSA reduce placental parasitaemia and improve pregnancy outcomes. Currently, no single assay is able to predict if a woman has adequate immunity to prevent placental malaria (PM). This study measured Ab levels to 28 malarial antigens and used the data to develop statistical models for predicting if a woman has sufficient immunity to prevent PM.MethodsArchival plasma samples from 1377 women were screened in a bead-based multiplex assay for Ab to 17 VAR2CSA-associated antigens (full length VAR2CSA (FV2), DBL 1-6 of the FCR3, 3D7 and 7G8 lines, ID1-ID2a (FCR3 and 3D7) and 11 antigens that have been reported to be associated with immunity to P. falciparum (AMA-1, CSP, EBA-175, LSA1, MSP1, MSP2, MSP3, MSP11, Pf41, Pf70 and RESA)). Ab levels along with clinical variables (age, gravidity) were used in the following seven statistical approaches: logistic regression full model, logistic regression reduced model, recursive partitioning, random forests, linear discriminant analysis, quadratic discriminant analysis, and support vector machine.ResultsThe best and simplest model proved to be the logistic regression reduced model. AMA-1, MSP2, EBA-175, Pf41, and MSP11 were found to be the top five most important predictors for the PM status based on overall prediction performance.ConclusionsNot surprising, significant differences were observed between PM positive (PM+) and PM negative (PM−) groups for Ab levels to the majority of malaria antigens. Individually though, these malarial antigens did not achieve reasonably high performances in terms of predicting the PM status. Utilizing multiple antigens in predictive models considerably improved discrimination power compared to individual assays. Among seven different classifiers considered, the reduced logistic regression model produces the best overall predictive performance.
KW - Predictive models
KW - Placental malaria
KW - Multiplex assays
KW - VAR2CSA
U2 - 10.1186/s12936-017-2041-3
DO - 10.1186/s12936-017-2041-3
M3 - Journal article
C2 - 28962616
VL - 16
JO - Malaria Journal
JF - Malaria Journal
SN - 1475-2875
M1 - 391
ER -
ID: 184768444