Convolutional neural network-derived neurofibrillary tangle classifiers: Investigative tools to identify maturation...

Area of Science:

• Neuroscience
• Biochemistry
• Artificial Intelligence in Medicine

Background:

• Post-translational modifications (PTMs) of tau protein are crucial in Alzheimer disease (AD) pathogenesis and serve as biomarkers.
• Conventional proteomics methods are too labor-intensive for detailed, histology-specific investigations of tau PTMs in brain tissue.
• Neurofibrillary tangles (NFTs) represent aggregated tau and exist in various maturation stages, from early pretangles to late ghost tangles.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI)-guided pipeline for laser capture microdissection (LCM) of NFTs based on their maturation level.
• To investigate the abundance and changes of specific tau PTMs across different NFT maturation stages using targeted proteomics.
• To assess the utility of AI-driven classification of NFTs for subsequent molecular analysis.

Main Methods:

• Machine learning algorithms were trained and evaluated to classify NFT maturation stages (pretangles, mature tangles, ghost tangles) in anti-pTau217 stained brain sections.
• The best-performing AI classifier guided LCM for harvesting specific NFT morphologic forms from 38 AD cases across two biobanks.
• Targeted mass spectrometry (MS) was used to quantify tau signature peptides (pTau181, pTau217, TauMTBR) in approximately 1250 collected NFT samples.

Main Results:

• AI classifiers achieved moderate performance (e.g., Classifier A: precision/recall/F1 of 0.6/0.46/0.5) in subtyping NFTs.
• Targeted proteomics revealed a significant increase in tau peptide abundance from pretangles to mature tangles (2-11-fold).
• A dramatic decrease in tau peptide levels was observed in ghost tangles (3-116-fold), with pTau217 showing the most pronounced change.

Conclusions:

• AI-guided LCM provides an objective method to enrich specific NFT morphologic forms for molecular investigation.
• The study demonstrates dynamic changes in tau PTMs during NFT maturation and clearance, offering insights into AD progression.
• This coupled LCM-MS approach, informed by pathologist-trained classifiers, enables detailed analysis of AD-associated PTMs in situ.