Extrapolating missing antibody-virus measurements across serological studies.

Area of Science:

• Virology
• Immunology
• Computational Biology

Background:

• Understanding viral evolution and host antibody responses requires measuring antibody-virus interactions.
• Current methods face challenges due to vast viral diversity and antibody repertoire complexity, making comprehensive testing infeasible.
• Existing studies often use limited panels and lack frameworks for combining data across different assays or host species.

Purpose of the Study:

• To apply matrix completion for predicting unmeasured antibody-virus interactions.
• To determine the number of measurements required to achieve accurate predictions.
• To demonstrate the ability to combine disparate datasets, even with limited measurements.

Main Methods:

• Application of matrix completion algorithms to large-scale influenza and HIV-1 antibody-virus interaction datasets.
• Analysis of prediction accuracy as a function of the number of measurements.
• Evaluation of dataset combination capabilities across studies with varying assay types and host species.

Main Results:

• Matrix completion accurately predicts antibody-virus interactions in low-dimensional spaces.
• Significant reduction in necessary measurements is achievable, potentially saving ~250,000 measurements in incomplete datasets.
• The method successfully integrates data from disparate studies, even below theoretical prediction limits.

Conclusions:

• Matrix completion offers a powerful and efficient approach to characterizing antibody-virus interactions.
• This method can reduce experimental costs and accelerate research in vaccine development and viral immunology.
• The approach is generalizable to other viruses and low-dimensional biological datasets.