Related Experiment Video
Updated: Nov 2, 2025
09:35
Analysis of Somatic Hypermutation in the JH4 intron of Germinal Center B cells from Mouse Peyer's Patches
Published on: April 20, 2021
6.9K
Machine learning analyses of antibody somatic mutations predict immunoglobulin light chain toxicity
Maura Garofalo1, Luca Piccoli1, Margherita Romeo2
1Institute for Research in Biomedicine, Università della Svizzera italiana, Bellinzona, Switzerland.
Nature Communications
|June 11, 2021
Summary
A new machine learning tool, LICTOR, predicts light chain (LC) toxicity in systemic light chain amyloidosis (AL). This approach aids in early diagnosis, potentially reducing organ damage and mortality rates associated with AL.
Area of Science:
- Hematology
- Computational Biology
- Medical Diagnostics
Background:
- Systemic light chain amyloidosis (AL) involves toxic light chain (LC) aggregates damaging organs.
- Delayed diagnosis is common in AL, often occurring after significant organ damage has occurred.
Purpose of the Study:
- To introduce LICTOR, a machine learning model for predicting LC toxicity in AL.
- To improve early diagnosis and reduce mortality in AL patients.
Main Methods:
- Developed LICTOR, a machine learning approach analyzing somatic mutation distribution in LC sequences.
- Validated LICTOR on independent LC sequences with known clinical phenotypes.
- Used in silico and in vivo (Caenorhabditis elegans) models to confirm LICTOR's findings.
Main Results:
- LICTOR achieved 0.82 specificity and 0.76 sensitivity, with an AUC of 0.87.
- The model demonstrated 83% prediction accuracy on an independent dataset.
- Reverting identified mutations in silico abolished LC toxicity, confirmed experimentally.
Conclusions:
- LICTOR offers a promising strategy for early AL diagnosis.
- The tool has the potential to significantly reduce AL-related mortality.
- Understanding LC mutation patterns can guide therapeutic strategies.
