6. Review Articles

Best Practices for Machine Learning-Assisted Protein Engineering: A review distilling practical guidance for applying machine learning to protein engineering campaigns, covering data curation, model selection, and experimental design/validation loops. The authors also maintain a companion “Protein Engineering Code Center” repository collecting useful tools and resources referenced in the review.
Published: December 8, 2025
Paper | Code Center (GitHub)

Generative AI for the Design of Molecules: Advances and Challenges: A review of generative AI methods for molecular design — VAEs, GANs, normalizing flows, and diffusion models — covering generative architectures, sampling/training/post-generation optimization strategies, and applications across small-molecule design (unconstrained and property-constrained), conformation modeling, and large biomolecule generation (proteins, antibodies, peptides). It surveys benchmarking datasets/metrics and translational case studies, notably AI-discovered antibiotics with in vivo efficacy against multidrug-resistant infections, and outlines open challenges around physics-model integration, data scarcity, and multi-objective optimization.
Published: November 18, 2025
Paper

A Comparative Review of Deep Learning Methods for RNA Tertiary Structure Prediction: A systematic evaluation of six deep learning RNA 3D structure predictors — DRfold, DeepFoldRNA, RhoFold, RoseTTAFoldNA, trRosettaRNA, and AlphaFold3 — on three benchmark sets: RNA-Puzzles, CASP15 RNA targets, and a newly compiled dataset of sequentially distinct RNAs for generalization testing (plus a fourth, more stringent set of sequentially and structurally distinct RNAs). No single model dominates across all targets, and the study also tests whether standard scoring functions (Rosetta score, ARES) can reliably pick the best model from a pool of predictions, and compares RoseTTAFoldNA vs. AlphaFold3 on RNA chains modeled in isolation versus as part of larger complexes.
Posted: December 3, 2024
Preprint