The units on all IgG measurements changed from IU/ML (in 2020) to MFI (in 2021 and 2022). For example, see the unit column in the raw data files “2020LD_plasma_ab_titer.tsv” and “2021LD_plasma_ab_titer.tsv”.
Does that mean that if we use the 2020 values to predict the 2021 values, we should expect some additional conversion factor between IU/ML and MFI, and if so, would this conversion be linear or more complicated? Alternatively, should we ignore the units and expect that the MFI columns should be directly comparable?
You’re right. The measurement units for antibody titer readouts shifted from IU/mL in 2020 to MFI in 2021 and 2022. In our first (internal) challenge, we attempted to predict raw MFI values from the 2021 dataset using the 2020 dataset’s IU/mL values. Unfortunately, our attempts using JIVE and MICE methods were unsuccessful in identifying any trends. This led us to normalize the plasma antibody titer dataset using the median of the baseline (day 0) values. Post-normalization, we observed much improved performance from both the JIVE and MICE methods.
Accordingly, for the second challenge, we have included ‘MFI_normalized’ values along with the raw IU/mL and MFI values for the antibody titer measurements. This approach ensures that antibody measurements across datasets are comparable.