How to Test Protein Similarity with Better Limits
Scientists employ hydrogen/deuterium exchange mass spectrometry to unravel how proteins fold.
When comparing two drug versions, the goal is a test that confirms samples are almost identical, not merely different.
The Traditional Approach
A conventional method, called TOST (Two One‑Sided Tests), sets acceptable difference limits.
Historically, these limits were derived by randomly shuffling eight repeated measurements from a reference protein.
Because each shuffle produces different limits, the test’s reliability fluctuated.
A New Strategy: Exhaustive Combinations
Researchers shifted to evaluating every possible combination of the reference data.
- Using all combinations, even those with large swings, yields more stable and typically larger limits.
- This reduces the influence of random variation on the test outcome.
Enhancing Precision: Three Ideas Tested
- Outlier Detection & Removal – Identifying and excluding anomalous data points.
- Percentile‑Based Cut Points – Setting limits based on specific percentiles of the data distribution.
- Data Splitting – Dividing the dataset into subsets and testing each separately.
Applying these refined methods to real drug samples—including three biosimilar versions of a cancer‑treatment antibody and a model protein—demonstrated:
- Accurate identification of truly similar samples.
- Correct flagging of dissimilar ones.
The improved approach offers a robust tool for regulatory reviews, ensuring drug consistency and safety.
