torchcell.datamodels.schema.MeasurementType

class MeasurementType(*values)

Bases: StrEnum

What an environment-response readout number IS, so heterogeneous chemogenomic scores are never silently compared.

• log2_ratio: log2(treatment/control) barcode abundance / fitness ratio (Vanacloig).

• z_score: standardized fitness/growth deviation (Wildenhain; Hoepfner z-score columns). It ALSO covers the variance-normalized colony-size score the E. coli chemical-genomics arrays release as a “fitness-score” (Shiver 2016, Nichols 2011 and the Collins 2006 pipeline they are computed with), and there is deliberately NO s_score member for that family. The reason is the sourcing rather than the statistics: no mirrored text calls the released number an S-score. Shiver 2016 says only that “These fitness-scores represent the statistical significance of a change in colony size for a particular condition, with negative and positive fitness-scores representing sensitivity and resistance, respectively”, and that its pipeline “built upon previous analyses [8,18]” – Nichols 2011 and Collins 2006, NEITHER of which is in the literature mirror (#691). Its own Methods name the standardizing step, “variance normalization of the data (to improve reproducibility of measurements between plates)”, which is what this member’s “standardized” means. A member named after a statistic whose definition we cannot read would be a label with no provenance behind it, so the S-score identification stays an explicitly labeled inference in the loader and in [[torchcell.datasets.ecoli.shiver2016]] (issue #749).

• sensitivity_score: HIP/HOP sensitivity / fitness-defect score (Hoepfner MADL, Hillenmeyer, Lee).

• categorical: NOMINAL qualitative call with no order among its terms (Auesukaree sensitive/tolerant; Mormino enhanced-signal/no-effect).

• ordinal: a RANKED grade on a source-defined scale, where the ranks are ordered but the spacing between them carries no meaning (Smith 2006 scores the clear zone 4/3/2/1 = larger than wild type / wild type / less than wild type / small or not detectable, and growth on acetate 3/2.5/2/1). Distinct from categorical because the order is real (a 2 lies between a 1 and a 3) and distinct from every numeric member because the number is a rank, not a measured quantity: averaging it or comparing it to a z-score is meaningless. Like categorical, it requires category; unlike it, environment_response may carry the rank itself.

• growth_rate: absolute or normalized growth rate / doubling time.

• relative_growth_rate: the RUN’s wild-type generation time divided by the well’s generation time, so 1.0 = grows like the wild type, below 1.0 = slower, above 1.0 = faster. Distinct from growth_rate, which is an absolute rate in the assay’s own units: this number is a dimensionless ratio against a wild-type control measured in the SAME run, so it is comparable across runs where a raw rate is not. Its denominator is the well, which is why a slower strain scores below 1.0.

• differential_fitness: SIGNED difference of normalized colony-size fitness in a test condition minus the matched reference condition (Costanzo 2021 condition-SGA: “the difference in colony size measured in a particular test condition versus the matched reference condition for each mutant”); negative = condition-hypersensitive, 0 = fitness unchanged vs reference. Distinct from growth_rate (a rate, which is non-negative) – a differential is routinely negative.

• control_regression_residual: SIGNED residual of a strain’s colony size in a test condition regressed on the SAME strain’s colony size on a matched control plate (Bloom 2019 segregant panels: residuals(lm(s.radius.mean ~ ctrl.s.radius.mean)) against the same-cross, same-batch YPD or YNB plate); 0 = grew exactly as predicted from control growth. Distinct from differential_fitness (a plain subtraction of two normalized fitnesses): a regression residual removes the control-plate slope, so the two are not comparable numbers.

• colony_size: ABSOLUTE end-point colony size (mean radius in image pixels for the Bloom 2019 control plates); non-negative, unnormalized, scale is the assay’s own.

• condition_growth_ratio: a strain’s end-point growth on the TEST plate divided by the SAME strain’s growth on the control plate of the same screen (Peter 2018: the “growth ratio between stress media and standard YPD medium at 30 C” of each of 971 natural isolates); 1.0 = grows as on the control, non-negative, 0 = no colony. Distinct from relative_growth_rate (the denominator there is the run’s WILD TYPE in the same condition, and the numbers are generation times), from differential_fitness (a signed difference, 0 at the control) and from control_regression_residual (a residual with the control slope removed).

• cell_yield: ABSOLUTE number of cells a colony population yields over its growth curve, from calibrated colony images (De Chiara 2022’s Scan-o-matic “total cell yields (y)”, released in cells). Distinct from colony_size, which is an end-point size on an image scale (pixels): this is a cell count, and the source does not say whether it is the final population or the gain over the inoculum, so it is named for what the source calls it.

• relative_fitness: SIGNED fitness of one barcode lineage relative to the mean fitness of the pooled population it competed in, as a lineage-tracking estimator reports it from read-count trajectories (Hale 2024: PyFitSeq Malthusian fitness on a time axis in generations, then shifted per assay to a stated reference); 0 = grows like the reference. Distinct from log2_ratio (an end-point abundance ratio, not a fitted growth rate) and from relative_growth_rate (a ratio against a wild-type well, not a difference against a pooled population).

• optical_density: ABSOLUTE absorbance of a liquid culture read at a stated wavelength after a stated incubation (Baba 2006 Supplementary Table 3: each Keio mutant’s OD600 after 22 h in LB and 24 h and 48 h in glucose MOPS), on the plate reader’s own scale and unnormalized. Distinct from growth_rate (a rate fitted over a curve; this is one end-point reading), from cell_yield (a cell count, not an absorbance) and from colony_size (a solid-plate image measure).