|
F R E Q U E N T L Y A S K E D Q U E S T I O N S
|
|
SOME REMARKS
|
HN-NonLin is a program that we developed to analyze the results obtained from the new HRP2 assay. As all malaria in vitro drug sensitivity assays provide similar raw data (i.e. dose-response curves that follow a sigmoid curve) it works equally well for other assays (like WHO schizont maturation and isotopic assay). The major advantages are probably the fact that it is free and that it is simple to use. However, there is considerably more sofisticated software packages for the analysis of nonlinear regressions. If you have been using other software (like log-probit regression programs) and if you are happy with the results: don't change a winning team.
|
|
WHAT DOES THE PROGRAM DO?
|
It performs a nonlinear regression analysis to obtain individual effective (or inhibitory) concentrations (ECs). The model is a polynomial regression model with log-transformed drug concentrations. This model very closely follows the data obtained from the test (similar to interpolation), while adjusting for single outliers. Unfortunately no model is perfect when it comes to estimating biological parameters.
|
|
WHY NONLINEAR?
|
While the drug response of larger parasite populations (i.e. larger numbers of isolates or strains) may generally be linearised by log-transforming drug concentrations and probit-transforming the growth inhibition, single isolates rarely follow this pattern. Nonlinear regression analysis therefore offers the best method to obtain effective concentrations for individual parasite strains.
|
|
WHERE IS 0%?
|
Schizont maturation tests (WHO assay) will generally deliver a real 0% concentartion. Isotopic and colorimetric tests, however, always produce a background. The program therefore assumes that the lowest value is 0% and that higher concentrations of drug will give a similar result. If you have baseline data available (e.g. a sample taken before starting the culture) you may prefer to subtract this value from all your results (this may be especially valuable if you expect the parasites to show some degree of growth even an highest drug concentrations).
Set the value 'LOWEST Y=0' to TRUE in the 'OPTIONS' sheet if you want the lowest Y value automatically to be used as 0 (range of Y will be between lowest and highest value). Set the value to FALSE to leave your Y values untouched. Use TRUE if you expect your tests to reach 100% inhibition (preset). Set the value to FALSE if your tests will not produce 100% inhibition (in this case do not forget to subtract your baseline values).
|
|
WHY DO I GET 0% FOR Y EVEN THOUGH THE RAW DATA STILL SHOW A LOW VALUE?
|
To eliminate variations around the lowest values, a certain percentage of the control may be subtracted from all results. This will prevent the program from delivering EC99 values, which may be way above the real data (as many other programs do). By default this percentage is set to 0% (i.e. nothing will be subtracted). The value may be changed in the 'ADVANCED OPTIONS' sheet. If there is more than one data point at 0% then only the first one will be included in the model calculation. If the model predicts values below 0% they are set to 0%. Similarly if the model predicts values above 100% they are set to 100%.
|
|
WHAT HAPPENS IF THERE IS NO DATA POINTS AT CONCENTRATIONS BELOW IC50?
|
The program assumes that the control value is found at a concentration, which is calculated by taking the geometric mean of the lowest concentration and the control. This will prevent the program from delivering unrealistically low IC50 values if there is no data points below, while not affecting the results if there is sufficient data available. This can only give an estimate. Try to adjust the concentration of the drugs so that your first data point is below IC50.
|
|
WHY DOES THE REGRESSION LINE IN THE GRAPH SOMETIMES GO ABOVE 100% OR BELOW 0%?
|
The regression line in the graph ('Y predicted') is drawn by Excel and follows the data points of the predicted values. However, it is drawn using a different algorithm and may therefore not always follow exactly the mathematical model used to calculate EC-values.
|
|
IS THIS THE ONLY SOFTWARE I CAN USE TO EVALUATE MALARIA IN VITRO DRUG SENSITIVITY DATA?
|
Certainly not! There is a number of other in vitro programs available and several statistics packages also provide functions that can be used to analyse malaria in vitro data. So why HN-NonLin? Good question. I was never very happy with the results I got from some other programs, others were difficult to handle (e.g. those working in a DOS environment), others very expensive. So I started to experiment with my own software. First with log-probit analysis: very nice for cumulative data (>20 samples pooled) but due to the linearisation of the data not very well suited for individual results. The logical consequence was to use a nonlinear model. One option is interpolation. Interpolation gives results that very closely follow the raw data. However, it overrates outliers that may be found in almost any drug sensitivty assay. Nonlinear regressions, on the other hand, closely follow the raw data but may compensate for single outliers. The model used in HN-NonLin is therefore a polynomial (nonlinear) regression model.
|
|
WHERE DOES THE NAME 'HN-NONLIN' COME FROM?
|
The name is taken from the name of the working directory on my harddisk (I had another program in a directory called 'nonlin'). I thought the name IV-NonLin (it sounded logical to me, IV for in vitro). Unfortunately other people rather thought of 'intravenous' or of the number 4. Your suggestions are welcome.
|
Drug Resistance
HRP2
