Predicting and interpreting large-scale mutagenesis data using analyses of protein stability and conservation

Area of Science:

• Protein Science
• Genomics
• Biophysics

Background:

• Understanding single amino acid changes is vital for protein science.
• Large-scale experimental data on protein variants is needed.

Purpose of the Study:

• To predict the functional consequences of single amino acid substitutions.
• To investigate the mechanisms underlying loss of protein function.

Main Methods:

• Collected and analyzed experimental data for over 150,000 variants across 29 proteins.
• Employed biophysical calculations to predict changes in protein stability.
• Utilized sequence conservation analysis and machine learning models incorporating protein structure and sequence alignments.

Main Results:

• Sequence analysis provided more accurate predictions of variant effects than stability predictions.
• Approximately 50% of variants causing loss of function were attributed to stability effects.
• A machine learning model integrating structural and sequence data improved variant effect prediction and mechanistic interpretation.

Conclusions:

• Large-scale experimental variant effect data can yield generalizable insights into loss-of-function mechanisms.
• Combining sequence and structural information enhances the prediction of protein variant impacts.
• Stability changes are a significant, but not the sole, driver of functional consequences for amino acid substitutions.