Summary
The Triangular update from AlphaFold2 is dispensable for high-accuracy Structure prediction (1,2). This module from AlphaFold2/3 has a memory footprint that scales cubically with respect to the size of the protein. RosettaFold circumvented the need for this by instead passing the 3D structure alongside distogram/Multiple sequence alignments processing, and looking at pairwise residue distances (which scales quadratically). More recently, PairMixer dispensed entirely with the module without any loss of accuracy.
Figures
Ref (2)
1.
Baek M, Anishchenko I, Humphreys IR, Cong Q, Baker D, DiMaio F. Efficient and accurate prediction of protein structure using RoseTTAFold2. openRxiv; 2023. Available from: https://doi.org/10.1101/2023.05.24.542179
2.
Ouyang-Zhang J, Murugan P, Diaz DJ, Scarpellini G, Bowen RS, Gruver N, et al. Triangle Multiplication Is All You Need For Biomolecular Structure Representations. 2025; Available from: https://arxiv.org/abs/2510.18870