Duboue-Dijon, E. | Search bioRxiv

bioRxiv2026

Deep-learning predictions of biomolecular structures : persistent limitations and new horizons extended by explicit ion addition

The advent of deep learning-driven tools such as AlphaFold has revolutionized the prediction of biomolecular structures, offering unprecedented accuracy and accessibility for proteins, RNA, and their complexes. While these tools have demonstrated remarkable success in benchmarking competitions and enabled experimentalists to generate models with ease, their widespread use has also highlighted persistent challenges. These include difficulties in assessing model confidence, limitations in predicting transmembrane domains, nucleic acids, conformational diversity, and interactions with ions or ligands, as well as the tendency to misfold intrinsically disordered regions (IDRs). In this perspective, we critically evaluate the strengths and limitations of current AI-based structure prediction tools through illustrative examples with a particular emphasis on the impact of explicit ion modelling. We notably report how the explicit addition of a few potassium cations to the prediction of IDRs or G-quadruplexes can trigger massive conformational switches compared to "dry" predictions. On this basis, we suggest modelling sequences both "dry" and in the presence of explicit potassium cations as a simple, practical way to sample alternative conformations and to expose disordered regions that current predictors tend to over-fold. We discuss the importance of reporting confidence metrics in publications to avoid overinterpretation. Furthermore, we address the unique challenges of RNA structure prediction, where data scarcity and structural complexity limit the performance of both classical and deep learning methods. Our analysis underscores the need for continued methodological advancements, integration of complementary computational tools, and expansion of high-quality experimental datasets. TOC Graphic O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=111 SRC="FIGDIR/small/740587v1_ufig1.gif" ALT="Figure 1"> View larger version (16K): org.highwire.dtl.DTLVardef@e950e9org.highwire.dtl.DTLVardef@1bf22b3org.highwire.dtl.DTLVardef@17f5229org.highwire.dtl.DTLVardef@1eb2e9f_HPS_FORMAT_FIGEXP M_FIG C_FIG

biophysics↗

bioRxiv2025

Sticky salts: overbinding of monovalent cations to phosphorylations in all-atom forcefields

Phosphorylation is a major post-translational modification, which is involved in the regulation of the dynamics and function of Intrinsically Disordered Proteins (IDPs). We recently characterized a phenomenon, which we termed n-Phosphate collaborations (nP-collabs), where bulk cations form stable bridges between several phosphoresidues in all-atom molecular dynamic simulations. nP-collabs were found to be sensitive to the combination of forcefields and cation types. Here, we attempt to assess the physical relevance of these nP-collabs by evaluating the strength of the cation/phosphate interaction through osmotic coefficient ({phi}) calculations on the model [Formula] and [Formula] salts, using different classical forcefields for phosphorylations. All force-fields were found to overestimate the strength of the interaction to various degrees. We thus designed new parameters for CHARMM36m and AmberFF99SB-ILDN using the Electronic Continuum Correction (ECC) approach, which provide remarkable agreement for{phi} values for both cation types and over a range of concentrations. We provide a preliminary test of these ECC parameters for phosphorylations by simulating the sevenfold-phosphorylated rhodopsin peptide 7PP and comparing secondary chemical shifts to experimental data. Conformational ensembles resulting from the ECC-derived phosphorylated forcefields display both qualitative and quantitative improvements with regard to full-charge forcefields. We thus conclude that long-lasting nP-collabs are artifacts for classical forcefields born from the lack of explicit polarization, and propose a possible computational strategy for the extensive parameterization of phosphorylations. The presence of long-lived nP-collabs in simulations produced using classical forcefields is therefore a serious concern for the accurate modelling of multiphosphorylated peptides and IDPs, which are at the center of research questions regarding neurodegenerative diseases such as Alzheimers or Parkinsons.

biochemistry↗