Soft disorder regions (SDRs) are segments of a protein whose experimentally observed structure varies across alternative structure determinations of the same molecule: they are resolved (ordered) in some X-ray crystal structures but missing from the model or flagged with high B-factors (large positional uncertainty, i.e. conformational flexibility) in others. The LoRA-DR-suite/SoftDis study that introduced this operational definition proposed, as an open question, that such conditionally or partially disordered regions may contribute to human disease by disrupting disorder-mediated interactions, with intrinsically disordered, hybrid, and disorder-to-order transition proteins as the suggested objects of study. The open question is whether that proposal is true and, if so, through what mechanisms. Although the general link between intrinsic disorder and human disease is heavily established, and disease mutations are known to be enriched in intrinsically disordered regions and in conditionally folded (disorder-to-order) regions within them, no published work has assessed disease association for the soft-disorder class itself — regions defined by cross-structure variability, including flexible segments outside curated disorder annotations — nor disentangled any such association from the known enrichment in conditionally folded regions, nor attributed disease variants to disruption of interactions mediated by SDR ordering. Establishing the association and its mechanisms, or demonstrating its absence beyond what known disorder-disease links explain, is the open problem.

Background

Most proteins fold into a single stable three-dimensional structure, but intrinsically disordered regions (IDRs) do not: they remain flexible and ensemble-like in isolation while still performing functions, especially in signaling and gene regulation, often by binding partner molecules through short sequence motifs. When an IDR binds a partner it frequently folds into a defined structure at the binding interface — a disorder-to-order transition; short elements that fold upon binding this way are called molecular recognition features (MoRFs). Interactions that depend on such transitions are referred to here as disorder-mediated interactions. A further refinement of this picture is that many regions are only conditionally disordered: disordered under some conditions and ordered under others, depending on binding partners, small molecules, post-translational modifications such as phosphorylation, or the redox state of the cell. Reviews of conditionally disordered proteins emphasize that environmental cues can be the trigger for folding (Hausrath and Kingston, 2017), and large-scale analysis shows, for example, that redox-sensitive conditionally disordered regions are widespread in higher eukaryotes (Erdős et al., 2019). In a related development, AlphaFold2, the deep-learning method that predicts protein three-dimensional structures from amino-acid sequence together with a per-residue confidence score, was shown to assign confident structures to nearly 15% of human IDRs, and these confident predictions often correspond to the conditionally folded (bound or condition-specific) state rather than a permanently folded one (Alderson et al., 2023). Against this backdrop, Lombardi, Seoane and Carbone (2025) introduced the operational definition of soft disorder — the concept itself having been introduced earlier by Seoane and Carbone (2021) — by comparing alternative X-ray crystal structures of the same molecule: a soft disorder region (SDR) is a region that is ordered in some experimental determinations of the protein but is either absent from the deposited model (unresolved electron density, a hallmark of conformational disorder) or carried in high-B-factor (structurally flexible) form in other determinations of the same molecule. Alongside the definition they released SoftDis, a database of SDR annotations for roughly 500,000 PDB (Protein Data Bank) chains, and LoRA-DR-suite, a sequence-based predictor of both IDRs and SDRs. SDRs thus differ from classical IDR annotations: they are defined by experimentally observed structural variability across independent structures rather than by disorder annotations or sequence-based prediction, and they include flexible segments of otherwise well-folded proteins. The disease context is well developed for intrinsic disorder in general. The 'D2 concept' review established intrinsically disordered proteins as a major class of disease-associated proteins (Uversky, Oldfield and Dunker, 2008), and the subsequent literature on disordered proteins in cancer and neurodegeneration is extensive. At the level of mechanism, 21.7% of annotated missense disease mutations fall within IDRs; disease mutations in IDRs disproportionately cause local disorder-to-order transitions (a 1.7-2.7-fold increase over annotated polymorphisms and neutral substitutions) and affect predicted MoRFs more often than control mutations (Vacic et al., 2012). Furthermore, the buried interface cores of interactions mediated by IDRs are strongly enriched in disease missense mutations and depleted in neutral ones, directly supporting disruption of disorder-mediated interactions as a disease mechanism (Wong et al., 2020). Most closely adjacent to the present question, human disease mutations were found to be nearly fivefold enriched in conditionally folded IDRs — regions that fold upon binding or under specific conditions, identified via AlphaFold2 confidence within annotated IDRs — compared with IDRs in general (Alderson et al., 2023). A recent proteome-wide functional map of human IDRs likewise contains clusters enriched for proteins encoded by disease-risk genes (Pritišanac et al., 2026). What none of this established is the disease association of soft disorder as its own class: regions defined by cross-structure variability, which partially overlap but do not coincide with annotated IDRs or AlphaFold2-confident conditionally folded regions, since SDRs also capture flexible segments of proteins never curated as disordered. In their discussion, Lombardi, Seoane and Carbone (2025) therefore posed, as an explicitly unresolved question, whether and how soft disorder contributes to human diseases via disruption of disorder-mediated interactions, suggesting that study of intrinsically disordered proteins, hybrid proteins (proteins combining ordered domains with functional disordered regions), and proteins undergoing disorder-to-order transitions could clarify disease mechanisms and reveal therapeutic targets. The open content of the question is therefore the SDR-specific disease association and mechanism, not the general disorder-disease link, which is already well established.

Problem Statement

Establish whether soft disorder regions — regions of a protein that are ordered in some experimentally determined structures of that protein but disordered or flexible in alternative structures of the same molecule, as catalogued in the SoftDis database through missing residues and B-factors across alternative X-ray crystal structures — contribute to human disease through disruption of disorder-mediated interactions, and if so, by what mechanisms. Concretely, an answer must determine: (1) whether disease-associated genetic variation and disease-risk genes are enriched in soft disorder regions relative to appropriately matched controls, beyond the already-demonstrated enrichment of disease mutations in intrinsically disordered regions and in conditionally folded (disorder-to-order) regions within them; (2) which interactions mediated by soft disorder regions are affected by disease-associated variants, and which conditioning factors — binding partners, ligands, post-translational modifications, or cellular conditions — govern the ordering of those regions in health and disease; and (3) validation of representative mechanistic cases in which a disease variant acts through disruption of a soft-disorder-mediated interaction. A demonstrated absence of any SDR-specific disease association, beyond what the established disorder-disease literature explains, would equally answer the 'whether' and close the question in the negative direction. The problem is posed at its natural generality: no particular disease, tissue, protein class, or mechanism is excluded, and the interaction set and cases called for in (2) and (3) are to be read as representative rather than exhaustive.

References

  1. Gianluca Lombardi, Beatriz Seoane, Alessandra Carbone, 'LoRA-DR-suite: adapted embeddings predict intrinsic and soft disorder from protein sequences', Bioinformatics 41(Supplement_1):i439-i448, 2025. DOI: 10.1093/bioinformatics/btaf185, https://doi.org/10.1093/bioinformatics/btaf185
  2. Gianluca Lombardi, Beatriz Seoane, Alessandra Carbone, 'LoRA-DR-suite: adapted embeddings predict intrinsic and soft disorder from protein sequences', bioRxiv preprint, posted 2025-02-07. DOI: 10.1101/2025.02.03.636253, https://doi.org/10.1101/2025.02.03.636253
  3. Vladimir N. Uversky, Christopher J. Oldfield, A. Keith Dunker, 'Intrinsically Disordered Proteins in Human Diseases: Introducing the D2 Concept', Annual Review of Biophysics 37:215-246, 2008. DOI: 10.1146/annurev.biophys.37.032807.125924, https://doi.org/10.1146/annurev.biophys.37.032807.125924
  4. Vladimir Vacic, Phineus R. L. Markwick, Christopher J. Oldfield, Xiaoyue Zhao, Chad Haynes, Vladimir N. Uversky, Lilia M. Iakoucheva, 'Disease-Associated Mutations Disrupt Functionally Important Regions of Intrinsic Protein Disorder', PLoS Computational Biology 8(10):e1002709, 2012. DOI: 10.1371/journal.pcbi.1002709, https://doi.org/10.1371/journal.pcbi.1002709
  5. Eric T. C. Wong, Victor So, Mike Guron, Erich R. Kuechler, Nawar Malhis, Jennifer M. Bui, Jörg Gsponer, 'Protein-Protein Interactions Mediated by Intrinsically Disordered Protein Regions Are Enriched in Missense Mutations', Biomolecules 10(8):1097, 2020. DOI: 10.3390/biom10081097, https://doi.org/10.3390/biom10081097
  6. T. Reid Alderson, Iva Pritišanac, Đesika Kolarić, Alan M. Moses, Julie D. Forman-Kay, 'Systematic identification of conditionally folded intrinsically disordered regions by AlphaFold2', Proceedings of the National Academy of Sciences 120(44):e2304302120, 2023. DOI: 10.1073/pnas.2304302120, https://doi.org/10.1073/pnas.2304302120
  7. Andrew C. Hausrath, Richard L. Kingston, 'Conditionally disordered proteins: bringing the environment back into the fold', Cellular and Molecular Life Sciences 74(17):3149-3162, 2017. DOI: 10.1007/s00018-017-2558-1, https://doi.org/10.1007/s00018-017-2558-1
  8. Gábor Erdős, Bálint Mészáros, Dana Reichmann, Zsuzsanna Dosztányi, 'Large-Scale Analysis of Redox-Sensitive Conditionally Disordered Protein Regions Reveals Their Widespread Nature and Key Roles in High-Level Eukaryotic Processes', PROTEOMICS 19(6):1800070, 2019. DOI: 10.1002/pmic.201800070, https://doi.org/10.1002/pmic.201800070
  9. Iva Pritišanac, T. Reid Alderson, Đesika Kolarić, Taraneh Zarin, Shuting Xie, Alex Lu, Aqsa Alam, Abdullah Maqsood, Ji-Young Youn, Julie D. Forman-Kay, Alan M. Moses, 'A functional map of the human intrinsically disordered proteome', Proceedings of the National Academy of Sciences 123(21):e2604562123, 2026. DOI: 10.1073/pnas.2604562123, https://doi.org/10.1073/pnas.2604562123