Disease phenotyping is a burdening step when evaluating plant resistance due to the labor and time required. It becomes even tougher when the same plot or plant is phenotyped more than once.

In this work, we addressed this problem from a statistical perspective. Our main goals were:

- Determine the optimal number of phenotypic evaluations that should be assessed on a field plot to characterize quantitative disease resistance.
- Find the best method to combine multiple time-series datapoints into a single, accurate score that is suitable to use as response variable in further data-driven procedures such as GWAS or Genomic Selection (GS).

The Area Under the Disease Progress Curve (AUDPC) is usually applied to map a set of time-series datapoints into a single score, which represents the resistance of the field plot to a determinate pathogen. Although efficient, it is prone to bias in scenarios of sparse missing data and unbalanced trials. We designed three alternative algorithms (Angle, GDD50 and Maxvar) to cope with these suboptimal scenarios.

We validate these algorithms in the scope of an international breeding program by studying the predictability of algorithm’s scores on empirical and simulated data using Wheat as host and Fusarium Head Blight as disease.