Bio Rna Structure Covariation Analysis

Version Compatibility

Reference examples tested with: R-scape 2.0+, Python 3.10+

Before using code patterns, verify installed versions match. If versions differ:

• CLI: <tool> --version then <tool> --help to confirm flags
• Python: pip show <package> then help(module.function) to check signatures

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Covariation Analysis

"Is my RNA's conserved secondary structure real?" -> Measure whether the base pairs covary across an alignment more than phylogeny and base composition alone would produce, and whether the alignment has the power to detect such covariation.

• CLI: R-scape -s alignment.sto to test a given consensus structure
• CLI: R-scape --cacofold alignment.sto to build a covariation-supported structure de novo (CaCoFold)

The governing principle: covariation is the gold standard, but a negative needs POWER

A compensatory (covarying) mutation is the strongest possible evidence for a base pair: if two columns change together across evolution so as to PRESERVE pairing (a G-C in one species becoming A-U in another at the same two positions), that directly evidences the pair, and a structure conserved by covariation across a deep alignment beats any single-sequence thermodynamic prediction. R-scape (Rivas, Clements & Eddy 2017) tests whether observed pairwise covariation EXCEEDS a phylogeny-aware null, separating real structural covariation from the apparent covariation that phylogenetic correlation and biased composition produce on their own.

The load-bearing nuance is that "no significant covariation" is NOT automatically "no structure" -- it can mean the alignment lacks the POWER to detect covariation (too few sequences, or sequences too similar, so there is not enough variation to observe compensatory changes). R-scape estimates, for each pair, the probability it would be called significant if it were a true pair (its power), so the result is a THREE-way verdict, not pass/fail:

Reporting only "R-scape found 0 significant pairs" without the power context is the central misuse: a low-power negative says nothing about the structure. R-scape draws the low- vs high-power line at an explicitly arbitrary 10% mean alignment power (the sum of per-pair power over the number of base pairs; Rivas et al. 2020): below ~10%, treat a negative as "cannot infer."

How R-scape decides

R-scape computes a per-pair covariation statistic (the G-test by default, with average-product correction to remove background phylogenetic signal), builds a null distribution by simulating alignments under the inferred phylogeny and base composition, and assigns each pair an E-value. A pair is significantly covarying when its E-value is at or below the target (default 0.05). It reports the number of expected covarying pairs found, their positions, the inferred substitutions at each, and the per-pair power. Significance is judged against the phylogenetic null, so a raw "positive covariation score" is not enough -- only covariation ABOVE the null counts.

Test a given consensus structure

The input is a Stockholm alignment with a #=GC SS_cons line (the structure to test) -- e.g. an Rfam SEED, an RNAalifold consensus, or a hand-curated structure.

# -s evaluates the pairs in the alignment's SS_cons; -E sets the E-value target (default 0.05).
# --outdir keeps R-scape's outputs (.cov, .power, .sorted.cov, R2R .svg) out of the CWD.
R-scape -s -E 0.05 --outdir rscape_out alignment.sto

With -s, R-scape runs TWO tests: one on the pairs in the proposed SS_cons, and a separate one on all OTHER possible pairs -- so a significantly covarying pair OUTSIDE the proposed structure is evidence for a better or alternative fold, not just a yes/no on the given one. (A bare R-scape alignment.sto without -s tests all possible pairs as one set; -s is what scopes the primary test to the proposed structure.)

Outputs include <msa>.cov (covarying pairs: positions, score, E-value, substitutions, power), <msa>.power (power analysis), <msa>.sorted.cov, and an R2R .svg/.pdf diagram. Read the diagram by its legend: R-scape marks significantly covarying pairs distinctly from pairs that are merely structurally compatible and from pairs inconsistent with the covariation, so the highlighted pairs are the ones with evolutionary support. The header reports nseq, alignment length, average identity, and number of base pairs.

Build a covariation-supported structure de novo (CaCoFold)

When there is no trusted structure to test, let covariation drive the fold. CaCoFold (--cacofold, also accepted as --fold) maximizes the support from significantly covarying pairs and can include pseudoknots as additional structure layers.

# Predict a structure from the alignment's covariation; writes a CaCoFold .sto with a new SS_cons.
R-scape --cacofold -E 0.05 --outdir rscape_out alignment.sto

The CaCoFold structure is grounded in evolutionary evidence rather than thermodynamics alone, which makes it a strong consensus to seed a covariance model (ncrna-search) or to compare against a thermodynamic fold (secondary-structure-prediction).