The Promise and Peril of Chemical Probe Negative Controls

Abstract

Chemical probes are selective modulators that are used in cell assays to link a phenotype to a gene and have become indispensable tools to explore gene function and discover therapeutic targets. Chemical probe off-targets are a confounding factor as the observed phenotype may be driven by inhibition of an unknown off-target instead of the targeted protein. A negative control, a close chemical analog of the chemical probe that is inactive against the intended target, is typically used to verify that the phenotype is indeed driven by the targeted protein. Here, we compare the selectivity profiles of four unrelated chemical probes and their respective negative controls. We find that controls that chemically deviate from the probe by a single heavy atom can be inactive against up to 80% of known off-targets if the chemical modification has a charge-neutralizing effect. In such cases, a loss in phenotype upon treatment with the negative control may be driven by loss of inhibition of an off-target. To expand this analysis, we inspect the crystal structures of 90 pairs of unrelated proteins, where both proteins within each pair is in complex with the same drug-like ligand. We computationally estimate that in 50% of cases, methylation of the ligand (a simple chemical modification often used to generate negative controls) at a position that will preclude binding to one protein (the intended target) will also preclude binding to the other (the off-target). These results emphasize the need to select negative controls with care and profile both chemical probes and negative controls against diverse protein arrays to verify that off-targets of probes are also hit by negative controls. When available, a best practice should be to verify that two unrelated chemical probes targeting the same protein elicit the same phenotype.

Will the chemical probes please stand up?

“…two probes of different chemotypes and two negative controls, also matched as different chemotypes). 41 …”

“…two probes of different chemotypes and two negative controls, also matched as different chemotypes). 41 5. There are many selective and potent compounds appearing in the recent medicinal chemistry and chemical biology literature that, while not officially yet declared in probe sources, include sufficient characterisation for useful probe criteria scoring.…”

“… 16,40 The evaluation of probes at scale involves comparing public data at different stringencies according to availability. The seven key criteria are: 1) <100 nM target potency in vitro , 2) <1 uM for cell-based assays with evidence of direct target engagement, 3) target selectivity >100-fold, 4) absence of structural alerts indicating chemical liabilities 5) identification of an inactive analogue as control with significantly lower potency or inactive against the primary target, 41 6) an orthogonal probe with a different chemotype against the same primary target, and 7) SAR data to increases confidence in specific target modulation. The simplest approach to assigning a compound as a probe is to use these criteria (or a subset thereof).…”

Will the Chemical Probes Please Stand Up?

“…two probes of different chemotypes and two negative controls, also matched for different chemotypes). 55 5. There are many selective and potent compounds appearing in the recent medicinal chemistry and chemical biology literature that, while not officially yet declared in probe sources, include sufficient characterisation for useful probe criteria scoring.…”

“…The calculated scores employ accepted probe criteria 53,54 applied at the stringency thresholds of a) < 100 nM target potency in vitro b) < 1 uM for cell-based assays (1 µM), and c) target selectivity > 30-fold. Other important criteria included d) structural alerts, e) identification of an inactive analogue control with significantly lower potency or inactive against the primary target in vitro 55 , f) an orthogonal probe with different chemotype against the same primary target or g) SAR data sets that increase the confidence of specific target modulation. The PMIS, ranging from 0 to 1, combines partial scores for 1) potency, 2) selectivity 3) activity in cells, 4) SAR data 5) availability of an inactive analogue and 6) a PAINS score.…”