Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2034
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-20T18:14:19.189795Z
status: published
lens: g_space
sector: science
headline: SPROTify: A protein-centric machine learning framework for small protein identification with validation in cancer proteomics
dek: Small proteins, typically under 100 amino acids in length, constitute an emerging class of functional biomolecules involved in diverse biological processes. Despite their importance, they are frequently overlooked in genome annotation, as conventional prediction tools prioritize canonical-length open reading frames (ORFs) or rely on nucleotide sequences that may not capture the protein-level physicochemical properties governing stability and detectability. To address this, we developed SPROTify, a protein-centri…
gain_title: A machine learning classifier built on amino acid features predicted small proteins accurately and found 334 candidates in six cancer proteomics cohorts, 36 of them differentially expressed.
problem_title: (none)
trace_subject: (none)
gain_reading: A machine learning classifier built on amino acid features predicted small proteins accurately and found 334 candidates in six cancer proteomics cohorts, 36 of them differentially expressed.
gain_evidence: In six cancer proteomics cohorts, 334 small-protein candidates were identified, including 36 differentially expressed small proteins
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built SPROTify, a machine learning framework that predicts small proteins, those under about 100 amino acids, from amino acid-derived features instead of nucleotide sequences. It was trained on curated, experimentally validated human data and tuned through hyperparameter optimization.

Small proteins are often missed by genome annotation tools that favor canonical-length open reading frames, so a protein-level approach could help prioritize them for study. The text describes functional roles in tumor biology only as suggested by structural modeling, so biological function and clinical value remain unconfirmed.
limitation: The structural modeling only suggests specialized functional roles in tumor biology; the abstract reports computational candidates and does not confirm their function.
tag: Evidence-backed gain
key_points: On an independent test set the model reached accuracy and F1-score of 0.92 and an AUC of 0.97. | Generalizability was assessed on cross-species, alternative-transcript/isoform-associated small-protein, and cancer proteomics datasets. | Six cancer proteomics cohorts yielded 334 small-protein candidates, including 36 differentially expressed small proteins with strong cancer-type specificity. | All scripts and datasets are freely available on GitHub.
rundown: The authors argue conventional prediction tools prioritize canonical-length open reading frames or depend on nucleotide sequences, which may miss protein-level physicochemical properties that govern stability and detectability.

After training on human data, the team tested the model on independent cross-species, isoform-associated and cancer datasets, reporting strong and consistent classification performance across these external validation sets.

Structural modeling of the 36 differentially expressed small proteins revealed unique fold characteristics and tissue-specific patterns, which the authors say suggest specialized functional roles in tumor biology.
sources:
- peer_reviewed | New Biotechnology | https://doi.org/10.1016/j.nbt.2026.10.006 | 2026-10-13
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

56f5d6ea413ce11d233506dd067e1165b4db546dd4a63a35fe552710a641d46b

previous

0000000000000000000000000000000000000000000000000000000000000000