We describe existing platforms for protein/peptide-based ligand identification and the drug delivery systems that might be exploited for the delivery of biologic-based degraders."
Link: pubs.acs.org/doi/10.1021/...

-- aim to approximate soft optimal denoising processes (a.k.a. policies in RL) that combine pre-trained denoising processes with value functions serving as look-ahead functions that predict from intermediate states to terminal rewards. "

- "We review these methods from a unified perspective, demonstrating that current techniques -- such as Sequential Monte Carlo (SMC)-based guidance, value-based sampling, and classifier guidance

- Construct full-length proteins with binding motifs and refining structures using the Rosetta FastDesign protocol and grafting (with a potential round of LigandMPNN optimization)
- Engineer and validate binders for Bcl2–venetoclax, DB3–progesterone, and PDF1–actinonin through experimental testing

- Benchmark MaSIF-neosurf against RFAA on 14 ligand-induced PPI complexes with 8,907 decoys from PDBBind
- Use MaSIF-search to predict buried surfaces and identify complementary surface fingerprints from a database of protein fragments (~640,000)

- Train sequence based models to predict the activity of regulatory elements (MPRALegNet, MPRAnn, EnformerMPRA, and SeiMPRA)
- Use MPRALegNet predicts TFBS combinations, fine-mapping and variant effects

Integrating genetic algorithms and language models for enhanced enzyme design
academic.oup.com/bib/article/...

- "This review systematically summarizes recent advances in chromatin interaction matrix prediction models...This article details various models, focusing on how one-dimensional (1D) information transforms into the 3D structure chromatin interactions"

EnzymeCAGE: A Geometric Foundation Model for Enzyme Retrieval with Evolutionary Insights
www.biorxiv.org/content/10.1...
Semantic mining of functional de novo genes from a genomic language model
www.biorxiv.org/content/10.1...

Bridging Sequence-Structure Alignment in RNA Foundation Models
arxiv.org/abs/2407.11242
Mapping targetable sites on the human surfaceome for the design of novel binders
www.biorxiv.org/content/10.1...

NeuralPLexer3: Physio-Realistic Biomolecular Complex Structure Prediction with Flow Models
arxiv.org/abs/2412.10743
FlowDock: Geometric Flow Matching for Generative Protein-Ligand Docking and Affinity Prediction
arxiv.org/abs/2412.10966

IgDesign: In vitro validated antibody design against multiple therapeutic antigens using inverse folding
www.biorxiv.org/content/10.1...

- Use gradient-based approximation to modify protein sequences to increase/decrease specific concept values (e.g., which amino acids for increasing aromaticity).

- Train model with MLM Loss, Concept Loss (mean square error on concept embedding), and Orthogonality Loss (cosine similarity between known/unknown embeddings).