AI Science Tracker

Authors in Science (see paper)

3 discoveries from 2022

Discoveries

Neural networks decode CAR costimulatory motif grammar and guide receptor design

Source check: Needs source check. Not independent scientific confirmation.

Finding: A library of ~2,300 synthetic CAR costimulatory domains from 13 motifs steered T-cell phenotypes; neural networks extracted design rules, e.g. non-native TRAF and PLCG1 motif combinations enhanced cytotoxicity and stemness. Abstract-level review only; in vivo efficacy not read.

Technology: Neural networks trained on library phenotypes

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Models decoded the combinatorial signaling grammar to propose CAR designs.
Test and result
Library screening with phenotypic readouts; model-guided designs tested.
Limitation / still unknown
Clinical benefit and in vivo breadth not assessed here.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/search?query=DOI%3A10.1126/science.abq0225

Needs source check · editorial details

The source has not been fully checked or contains conflicting information. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review scope and date are not recorded; source checked does not establish full-text review.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 3/5 · Validation 3/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Network hallucination generates symmetric protein assemblies confirmed by structure

Source check: Needs source check. Not independent scientific confirmation.

Finding: Deep-network hallucination generated symmetric homo-oligomers from protomer number and length; crystal structures of seven designs matched models (median RMSD 0.6 angstrom) and three cryo-EM structures of giant rings up to C33 symmetry were obtained. Abstract-level review only.

Technology: Deep network hallucination (AlphaFold/RoseTTAFold-class)

Family AlphaFold: profile · models page

Publication: Peer-reviewed paper · Test context: prospective or independent check. Catalog categories, not an evidence grade.

AI contribution
Predictor-driven hallucination produced new backbones and sequences.
Test and result
Seven crystal and three cryo-EM structures compared with models.
Limitation / still unknown
Fraction of failed designs not read; full text not read.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/search?query=DOI%3A10.1126/science.add1964

Needs source check · editorial details

The source has not been fully checked or contains conflicting information. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review scope and date are not recorded; source checked does not establish full-text review.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 4/5 · Validation 4/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Constrained hallucination and inpainting scaffold protein functional sites

Source check: Needs source check. Not independent scientific confirmation.

Finding: Two deep-learning approaches (constrained hallucination and RoseTTAFold inpainting) design scaffolds around specified functional sites, producing candidate immunogens, receptor traps, metalloproteins, enzymes and binders, validated with in silico and experimental tests. Abstract-level review only; success rates per design class not read.

Technology: RoseTTAFold-based constrained hallucination and inpainting

Family RoseTTAFold: profile · models page

Publication: Peer-reviewed paper · Test context: direct test. Catalog categories, not an evidence grade.

AI contribution
Trained structure-prediction networks generate sequences and structures that embed a target site.
Test and result
Designs checked in silico and experimentally (details in full text).
Limitation / still unknown
Per-class experimental success rates and function levels not assessed here.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://europepmc.org/search?query=DOI%3A10.1126/science.abn2100

Needs source check · editorial details

The source has not been fully checked or contains conflicting information. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review scope and date are not recorded; source checked does not establish full-text review.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 4/5 · Validation 3/5 · Novelty 4/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.