Functional Testing and Mutation Analysis
Why Is This Important?
Sequencing identifies numerous variants that differ among individuals, yet a significant gap exists between the mere discovery of a variant and the conclusion that it causes disease. Particularly for rare variants with limited observational cases, determining their significance based solely on population frequency or familial data is challenging. In such instances, functional assays provide critical evidence by directly interrogating how a variant alters protein or cellular processes.
However, “functional alteration” and “disease causation in patients” are not equivalent statements. Functional assays measure the effect of a variant within a specific experimental system and readout. Clinical interpretation requires an evaluation of how well that readout represents the actual disease mechanism.
Core Concepts
Variant interpretation aggregates independent lines of evidence, including population frequency, computational predictions, patient phenotype, segregation within the family, de novo status, functional assays, and other clinical observations. The ACMG/AMP framework classifies and combines this evidence into pathogenic and benign directions. Functional assays constitute one such line of evidence; well-validated assays may receive greater weight.
Functional assays vary widely. Different readouts can be employed, including enzyme activity, protein stability, subcellular localization, splicing, transcriptional activity, and cell viability. Because the biological layer measured by an assay may differ even for the same variant, results may vary; therefore, one must focus on what and how the assay measures rather than merely its name.
The Scale Transformed by MAVE
MAVE (Multiplexed Assays of Variant Effect) generates hundreds or thousands of variants in parallel to measure functional scores. This approach offers the advantage of mapping the effects of many potential variants prior to their discovery in patients. While it enables broad comparisons under consistent conditions compared to single-variant experiments, clinical validation is not automatically achieved simply due to scale.
MAVE scores represent the 'relative functional effect observed in this assay.' It is necessary to evaluate how well they distinguish known pathogenic and benign controls, whether biological replicates are reproducible, whether the score range is unsaturated, and how the median is interpreted. Thresholds for clinical classification require independent validation rather than being selected post hoc based on the data.
Small Example
Assume the disease mechanism of a gene is loss of function. If enzyme activity of the variant protein is nearly abolished in repeated assays and normal/pathogenic controls are well separated, this may serve as functional evidence for pathogenicity. However, if the disease arises via gain of function, or if the cell system used fails to recapitulate conditions of the relevant tissue, the interpretation of such results becomes less robust.
Conversely, a readout within the normal range does not exclude abnormalities in functional domains not assessed by that assay. This is why a single normal result cannot be used to declare benignity.
Questions to Ask for Evidence of a Positive Assay
- Is the assay readout linked to a known gene–disease mechanism?
- Are normal, benign, and pathogenic controls and replicates included?
- Was the evaluation performed by an experimenter blinded to the variant class?
- Are the dynamic range and technical variability adequately reported?
- Are the results reproducible across other experimental systems or clinical observations?
- Are the transcript, cell line, construct, and analytical threshold explicitly specified?
Relationship to Computational Prediction
In silico prediction prioritizes candidates using conservation and sequence/structural features but does not measure actual function. Do not treat the consensus of multiple correlated tools as if they constituted multiple independent lines of evidence. Verify version, training data overlap, calibration, and applicable variant types, and document these distinctions separately from functional assays and clinical data.
Limitations
Cell lines and artificial expression systems simplify the developmental stage, expression levels, and interactions of patient tissues. Large-scale studies may encounter missing data and batch effects, and cutoffs not defined a priori may be overfitted. Most importantly, loss of function is not a sufficient condition for pathogenicity, nor are normal readouts a sufficient condition for benignity.
Final classification must situate functional results within the patient’s phenotype, mode of inheritance, and other independent evidence. This article does not replace clinical classification or diagnosis of specific variants.
Reading in Connection
The method for distinguishing phase and individual pathogenicity when two variants are observed together is continued in the Compound heterozygous Concept. The reason protein structure prediction cannot replace functional assays is connected to the AlphaFold confidence and limitations Concept.