AI-Assisted Molecular Review Refines SMO Inhibitor Candidacy in SHH Medulloblastoma.
Chen Buxbaum, Anat Yahav-Dovrat, Yael Fisher et al.
✦ AI-curated · Sources linked
Case in brief
A 40-year-old woman with unresectable SHH-activated medulloblastoma received standard chemoradiation, but post-treatment imaging was equivocal. NanoString classified the tumor as WNT, whereas methylation profiling and pathology supported SHH activation; a co-occurring TP53 mutation placed her in a very-high-risk category. AI-assisted integration of pathology, methylation, sequencing, clinical, and imaging data favored the SHH classification, corroborated by GAB1 and beta-catenin staining. Sequencing found no PTCH1, SMO, or SUFU mutation, leaving SMO-inhibitor response indeterminate. No inhibitor was given; close surveillance was chosen. She remained in remission, with 11-month progression-free survival at last follow-up.
What made this case unusual
The case combined discordant molecular classification with equivocal imaging; AI-assisted synthesis reconciled the findings and framed the absence of canonical pathway mutations as indeterminate evidence for SMO-inhibitor response, rather than proof of no potential benefit.
Diagnostic / clinical pearl
When molecular classifiers conflict, integrate methylation, pathology, immunostaining, sequencing, and clinical context; in this case, methylation supported by GAB1 and beta-catenin staining favored SHH activation over the NanoString WNT result.
Why it matters
The case illustrates how AI-assisted review may help organize complex, discordant data in a rare tumor, while the treatment decision remained uncertain and was verified against specialist readings.