Research Library

Your saved research

Papers, articles, reports, and filings you've saved — each with an AI-generated recap so you never lose context.

9 items
Type
Tags
PaperSaved 3d ago
All-atom de novo design of protein binders against small-molecule targets

Krishna, Baek, Watson, Bennett et al. (Baker lab) · Nature 634, 2025

AI recapIntroduces RFdiffusion-AA, an all-atom extension of RFdiffusion capable of designing protein binders that recognise small-molecule ligands with atomic accuracy. Reports ≥20% success rate against 5 diverse targets, competitive with evolved natural binders.
“Revisit before writing next LinkGevity binder post.”
RFdiffusionDavid Bakerbinder design
PaperSaved 1w ago
Boltz-2: improved affinity prediction via calibrated confidence head

Chai Discovery team · bioRxiv 2026.02.14

AI recapReplaces Boltz-1's implicit affinity output with a calibrated confidence head trained on PoseBusters + merged BindingDB/ChEMBL. 18% accuracy gain on held-out PoseBusters; inference latency unchanged. Weights released under MIT.
Boltz-2docking
ArticleSaved 5d ago
AlphaFold3 is changing how pharma writes RFPs — quietly

Endpoints News

AI recapInterviews with four pharma BD leads show that AlphaFold3 access is now a standard line-item in CRO and biotech partnership contracts. Shift from “do you have comp chem?” to “do you have AF3 + a fine-tuning pipeline?” Framework for negotiating access covered.
AlphaFold3deal
ReportSaved 2w ago
Q1 2026 AI drug-discovery deal flow

biotech.today

AI recap58 publicly disclosed deals in Q1, down 12% from Q4 2025 in count but up 22% in total value. Top modalities: PROTAC (18), ADC (11), cell therapy (8). Median pre-money for AI-native Series A raised to $60M from $42M a year ago.
dealPROTAC
FilingSaved 4d ago
Arvinas 8-K: VERITAC-2 Phase 3 topline results

SEC EDGAR

AI recapStatistically significant PFS improvement (HR 0.64) vs. fulvestrant in HR+ metastatic breast cancer. First oral PROTAC to hit Phase 3 primary endpoint. Management commentary suggests NDA submission Q3 2026.
“Key comparator for our Q3 strategy review.”
ArvinasNCT05501886PROTAC
PaperSaved 3w ago
Calibrating ADMET predictions for rare chemotypes via few-shot fine-tuning

Vats, Karpathy, Barzilay · J. Chem. Inf. Model. 2026

AI recapProposes an active-learning loop where a general ADMET model is fine-tuned on 20–80 assayed compounds before screening unseen chemistries. Reduces false-positive rate on natural-product and macrocycle libraries by ~45%. Code open-sourced.
ADMET-AIADMETRegina Barzilay
PaperSaved 1w ago
Generative antibody design hits the developability wall

Fleishman, Bahl et al. · Nature Biotechnology 2026

AI recapBenchmarks 6 generative antibody design pipelines against 3 target classes. All methods produce binders; only 14% clear a baseline developability panel (aggregation, PK, immunogenicity). Calls for joint binding + developability training objectives.
DiffAbSarel Fleishmanantibody design
PatentSaved 2w ago
US 2026/0012345: PROTAC-based E3 ligase recruiters for CDK4/6

USPTO filing

AI recapArvinas-assigned application covering a new VHL-recruiting linker class for CDK4/6 degraders. Claims include composition + method of treatment in HR+ breast cancer. Continuation of 2023 priority filing.
ArvinasPROTAC