Meffil - coding of unknown phenotypes

[janful]

Hi there. We are currently using meffil to run an EWAS in some Illumina methylation data from ~200 subjects who are part of a longitudinal study, and are trying to run an analysis in a subset of participants who have subsequently converted to disease status versus participants who have definitely remained disease free (the remainder having no data or are ambiguous as they have developed some associated symtoms). We are running into issues in relation to getting the program to run in the reduced dataset (n=58 with longitudinal outcomes of interest). I imagine this is because there are not sufficient samples across each sentrixID from the whole experiment (and other user defined covariates) to allow computation in just 58 subjects. We have tried to include all 212 subjects, coding those individuals with ambiguous phenotypes as "-9" to indicate missing data so that all sentrixIDs are included in input, but the software seems to think this "-9" value is an actual category and is computing differences between the three coded groups. Similarly it fails when "NA" is included as the missing value. Is there a way we can include all data from 212 subjects as input, but only phenotype coding for the 58 participants with known disease outcomes? We have tried including 212 subjects in covariates file, and only 58 subjects in pheno file, but this also fails to run.
Any advice you can provide would be much appreciated.