A mathematical preprint reports a theorem covering every target in a narrow class of RNA secondary structures: those with at most two maximal helices made of two base pairs. In the paper's four-letter Watson-Crick model, each qualifying target is shown to have a sequence for which that target is the unique compatible, noncrossing structure with the maximum possible number of base pairs. Here, designable means meeting that mathematical condition, not demonstrating a sequence's behavior in a biological setting.
The result is a sufficient certificate for an existing modulo-2 procedure, using parity to distinguish the intended pairing pattern from alternatives within the model. The authors describe the work as a global resource-counting extension of that framework, while stating that it is not a new general decision capability or an improved asymptotic bound.
A tightly drawn mathematical class
The target class, labelled K≤2, is tightly defined. It allows at most two maximal helices of length two, excludes maximal helices of length one, and requires every other maximal helix to contain at least three base pairs. The theorem quantifies over every target that satisfies those structural rules, rather than making a blanket claim about RNA structures outside the class.
The underlying pairing system is deliberately spare. It uses four letters and permits only A-U and C-G pairs. Structures are pseudoknot-free, adjacent positions may pair, and lower energy is defined as having more pairs. The theorem therefore addresses a maximum-pair combinatorial problem, not the full physical process by which RNA adopts a shape.
The coloring step assigns black, white and gray labels to parts of the structure. A proper uniform modulo-2 separated coloring puts all unpaired nodes in one parity class and all gray paired nodes in the other. That pattern becomes the certificate used in the sequence-construction step.
How the proof handles short helices
The proof is constructive rather than merely existential. It proceeds by induction over helix subtrees and tracks finite sets of allowed entry states. The set F applies to subtrees with no isolated stack, while Q covers subtrees containing one or two isolated stacks. At an η entry, Q requires a gray first pair; F also permits a black or white first pair.
The critical counting step comes when two child subtrees contain isolated stacks. A long incoming helix can reach η while ending non-gray, allowing both child helices to start gray. In the paper's terms, that resolves two downstream short-helix demands within the stated target class.
Once the tree and helix representations and the deterministic choices are fixed, the recursive construction is reported to use linear time and linear space. The Lean development establishes existence and correctness, but it is not presented as extracted executable code.
Formal checks with a clear boundary
The complete theorem is reported as formalized in Lean 4 using Mathlib. The reported transitive axiom set is limited to propext, Classical.choice and Quot.sound.
The frozen source was rebuilt in an isolated environment. The qualification rebuild completed 3,070 jobs and checked dependencies, unfinished proof steps and source preservation. These checks support the reported reconstruction of the formal artifact, but they do not establish biological realism or replace independent human expert review.
The theorem excludes wobble pairs, pseudoknots, stacking and loop energies, ionic effects, ensemble objectives, kinetics and three-dimensional constraints. It therefore does not show that the constructed sequences remain uniquely optimal once those omitted biological and physical effects are included.
The paper also treats the three-isolated-stack case as a boundary for this modulo-2 certificate only. A certificate can fail in that setting without implying general nonseparability or undesignability, and the example is described as remaining ordinarily separated and designable.
What the preprint says about its status
The supplied manuscript is identified as arXiv:2608.25194v1 and dated 25 August 2026. At that version, it reports foundational and pervasive generative-AI assistance under the author's direction and says that completed independent human subject-matter-expert review had not yet occurred.
The paper reports a version 1.0.3 archival release containing the manuscript, Lean materials, example-checking output, audits and qualification artifacts, with a canonical repository and archival DOI identified. The author states that the work was undertaken through Szilard Scientific, LLC, in a private capacity, and that the views are the author's own rather than those of an employer or client.
Paper data and sources
Original title: Designability of RNA Targets with Up to Two Length-2 Helices
Authors: Ashutosh S. Jogalekar
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-25
DOI: Not available
Original paper · Full text