The inference code currently still uses "strain" as the basic sample identifier. This should be replace universally with sample_id.