Assess robustness to broader calibration errors and correlated probe calls

Investigate whether larger or lopsided errors in the calibrated probe-response matrix, or correlations among probe calls, compromise ProteoEM abundance estimates, beyond the uniform calibration shift evaluated in the simulations.

Background

ProteoEM uses a fixed, externally calibrated emission matrix of probe-response probabilities to estimate proteoform abundances from affinity traces. In the simulations, a particular shared calibration perturbation—a uniform logit shift with added noise—had little effect on the abundance estimates. However, the analysis did not establish robustness to other forms of calibration error or to dependence among probe calls.

The unresolved issue is whether asymmetric or larger calibration inaccuracies, as well as correlated measurement errors between probes, can materially bias the likelihoods and consequently the inferred proteoform composition. Resolving this would be important for assessing the method on experimental data, where calibration errors and correlated noise may occur.