RNA Structure Scoring

How SPACER evaluates target site accessibility using RNA secondary structure predictions.

Why RNA Structure Matters

RNA molecules fold into secondary structures through intramolecular base pairing. When a target site is buried within a stable hairpin or stem-loop, the Cas13-crRNA complex must first unfold the local structure before it can bind — reducing effective cleavage activity. Guides targeting accessible (single-stranded) regions of the target RNA consistently outperform those targeting highly structured regions.

Structure scoring is especially important for Cas13 because the target is single-stranded RNA, which folds extensively. For Cas12 (targeting dsDNA), secondary structure of the DNA is less relevant, though the crRNA itself can fold — SPACER currently focuses on target-side accessibility.

Prediction Method

SPACER computes the minimum free energy (MFE) of the local RNA structure around each target site using thermodynamic folding algorithms based on the Vienna RNAfold energy model. The MFE reflects how energetically favorable folding is for that region:

MFE (kcal/mol)StructureAccessibility
Near 0Little or no foldingHighly accessible — ideal
-5 to -15Moderate structurePartially accessible — acceptable
< -15Strong stable structuresPoorly accessible — likely reduced activity

Local Folding Window

Rather than folding the entire target RNA (which is computationally expensive and less informative for local accessibility), SPACER folds a window centered on each protospacer site. The window extends beyond the protospacer on both sides to capture flanking structure context:

text
Target RNA:  ...NNNNNN[   protospacer (S nt)   ]NNNNNN...
                    |←  flank  →|                 |←  flank  →|
                    └──────── folding window (~50–80 nt) ────────┘

The flanking length is configurable, but the default of ~15 nt on each side captures the most relevant structural context without introducing noise from distant structures.

Scoring Function

The raw MFE is converted to a 0–1 score using a sigmoid-like mapping. Sites with MFE near zero (unstructured) receive scores approaching 1.0, while heavily structured sites receive scores approaching 0.0:

MFE RangeScoreInterpretation
> -3 kcal/mol0.9–1.0Highly accessible target site
-3 to -100.5–0.9Moderately accessible — acceptable
-10 to -200.2–0.5Partially structured — may reduce activity
< -20 kcal/mol0.0–0.2Strongly structured — poor accessibility

Computational Cost

RNA folding is the most computationally expensive scoring component. For each candidate guide, a separate folding calculation must be performed. SPACER mitigates this through:

  • Local folding: Only the window around each site is folded, not the full target
  • Parallel computation: Multiple folding calculations run concurrently across available cores
  • Optional enablement: Structure scoring can be disabled entirely for faster analysis
Warning
Structure scoring is disabled by default because it requires additional computation time. For long targets with thousands of candidate guides, enabling structure scoring may significantly increase analysis time.
Tip
If you enable structure scoring, consider running a first pass without it to identify your top candidates, then re-analyzing only those candidates with structure scoring enabled. This two-pass approach can dramatically reduce computation time for large targets.