RNA Drug Discovery · DSSR-Enabled · 1,098 Complexes

SPIRAL

Structural Pockets and Interacting RNA-Associated Ligands

A curated meta-analysis of 1,098 RNA–small molecule structures annotated with DSSR-computed tertiary interaction parameters — stacking geometry, hydrogen-bond topology, groove engagement, and binding pocket context — across six functional RNA categories.

1,098PDB structures

1,137binding events

6RNA categories

275affinity entries

Browse structures ↓ Open 3D viewer → CBQS scores ↓

scroll

Composite Binding Quality Score

CBQS: category-independent binding quality

A seven-metric framework (M1–M7) normalized across all 1,137 SPIRAL entries. Captures hydrogen-bond quality/density (M1/M2), stacking quality/density (M3/M4), backbone contact quality/density (M5/M6), and pocket structural complexity (M7). Equal scores can arise from mechanistically distinct strategies — the sub-score breakdown is required to distinguish them.

Overall CBQS (0–100 scale)

32.2

Riboswitches

Highest — minor groove / H-bond

19.4

Regulatory motifs

Lowest — surface docking

26.1 / 24.1

Aptamers / G-quad

Equal score, opposite strategy

28.9

Ribozymes

Balanced across all 7 metrics

Sub-score profiles (M1–M7)

Category M1M2M3M4 M5M6M7

Riboswitches 34.034.223.647.9 21.023.541.1

Ribozymes 32.332.618.337.5 20.222.738.4

Syn. aptamers 21.717.438.453.3 13.411.726.8

G-quadruplexes 8.76.843.558.1 8.87.035.6

Ribosome RNA 31.323.512.819.4 21.418.933.4

Regulatory motifs 15.19.835.642.5 12.58.611.8

          M1/M2 H-bond quality/density · M3/M4 stacking quality/density
M5/M6 backbone contact quality/density · M7 pocket complexity
        

Design gaps

Five quantifiable gaps in regulatory RNA motif binders

Compared against the riboswitch benchmark, regulatory RNA motif binders show complete absence of junction loop contacts and severe underrepresentation of pseudoknot, multiplet, 2′-OH, and non-canonical contacts — despite ranking first in coaxial-stack contacts. Engaging these sites is predicted to improve both potency and selectivity simultaneously.

Design gapReg. motifsRiboswitchFold gap

Junction loop contacts0.0004.346∞

Pseudoknot contacts0.0150.86257.5×

Base multiplet contacts0.5974.9928.4×

Ribose 2′-OH contacts0.3582.3706.6×

Non-canonical contacts1.1495.4144.7×

Binding affinity analysis

C2′-endo and buried surface area predict affinity

Analysis of 275 non-redundant affinity entries (Kd, Ki, IC₅₀ → pKd scale) using Spearman rank correlation and cross-validated regression. C2′-endo enrichment marks induced-fit binding depth at junction loops and pseudoknots — the same sites most underengaged by regulatory RNA motif binders.

ρ = +0.530

C2′-endo pucker count vs pKd

n = 275 · p < 0.001 · strongest predictor

ρ = +0.440

Buried contact surface area vs pKd

n = 275 · p < 0.001 · second strongest predictor

ρ = −0.444

G-quadruplex H-bonds vs pKd

n = 30 · p = 0.014 · sign reversal vs riboswitch

Simpson's paradox — H-bond count

Global: ρ = −0.016 (n.s.) · Within riboswitches: ρ = +0.384 (p<0.001) · Within ribosome: ρ = +0.544 (p=0.04) · Within G-quadruplexes: ρ = −0.444 (p=0.014). H-bond optimization in one RNA class cannot be assumed to transfer to another.

Mean pKd by C2′-endo × BSA quadrant (n=275)

7.8

High C2′-endo
High BSA

tightest binders

5.8

Low C2′-endo
High BSA

size-driven

6.0

High C2′-endo
Low BSA

depth-driven

5.5

Low C2′-endo
Low BSA

weakest binders

G-quadruplex sub-finding

Within G-quadruplex entries, C2′-endo is the strongest within-category affinity predictor (ρ = +0.775, p < 0.001), indicating that the highest-affinity G-tetrad binders achieve potency through extended aromatic stacking rather than complex pocket engagement.

Analytics

Database statistics

Distribution of the 1,137 binding events across RNA category, experimental method, and binding pocket topology.

Binding events by RNA category

Experimental method

Binding pocket topology (cluster profile)

Top ligands by frequency

Per-entry annotations

What's included in every entry

⟳

Stacking geometry

ΔSASA-based quantification of RNA–ligand stacking. Per-nucleotide stacked atom detail. Distinguishes deep intercalation from tangential contact with the same nucleotide count.

🔗

H-bond topology

Each bond classified by RNA moiety (nucleobase, ribose 2′-OH, phosphate) and donor–acceptor atom pair type. Quality graded standard / acceptable / questionable / unknown by DSSR geometry criteria.

◎

Tertiary motif context

Every contacted nucleotide tagged with its DSSR structural element — stem, hairpin loop, junction loop, internal loop, pseudoknot, coaxial stack, base multiplet, non-canonical pair, G-tetrad.

◈

Groove engagement

Major/minor groove contact counts and groove-preference classification. Embedded flag assigned when the ligand SASA is near-completely reduced by RNA enclosure on both sides.

⊕

CBQS sub-scores

All seven CBQS metrics (M1–M7) for every entry. Category-level means, cluster assignment, and ranked position within the global SPIRAL distribution for benchmarking and virtual screening.

⟁

Affinity integration

275 non-redundant entries carry experimental pKd values from PDBbind + manual curation. Spearman correlation profiles included for all interaction parameters.

Data formats

Ready-to-use data formats

SPIRAL ships as per-PDB DSSR JSON files for full atomic detail and a flat CSV for quick pandas/Excel filtering. No database setup required.

per-PDB JSON (DSSR format)

str_id4-char PDB ID

categoryRNA functional class

clusterBinding mode cluster (C1–C6)

cbqsComposite Binding Quality Score

num_hbonds, with_base, with_sugar, with_po4H-bond counts by moiety

num_stacked_nts, dSASA_stacked_ntsStacking count and buried area

dSASA_contacted_residuesTotal buried contact surface area (Å²)

embedded, major_groove_binderTopology flags (Boolean)

ct_C2endoC2′-endo pucker count (affinity predictor)

contacts[], hbonds[], stacked_nts[]Full atomic detail arrays

flat CSV (spiral_summary.csv)

pdb_idStructure identifier

category, clusterRNA class and binding mode

cbqs, m1–m7All CBQS scores

ligand_tagCCD 3-letter code

mol_weightLigand MW in Daltons

pkdpKd (−log₁₀ Kd) where available

ct_C2endo, dSASA_contacted_residuesTop affinity predictors

embedded, minor_groove_binderBinding mode flags

python · filter by CBQS and affinity predictors

import pandas as pd, json, pathlib # ── Load SPIRAL flat CSV ─────────────────────────────────── df = pd.read_csv("spiral_summary.csv") # Filter: high C2'-endo + large BSA (top-quadrant affinity predictors) high_affinity = df[ (df["ct_C2endo"] >= 2) & # above median C2'-endo (df["dSASA_contacted_residues"] >= 582) & # above median BSA (Å²) (df["category"] != "Ribosome RNA") # exclude ribosome ] print(f"High-affinity pocket entries: {len(high_affinity)} structures") # Load per-PDB JSON for full interaction detail hits = [] for pdb_id in high_affinity["pdb_id"]: f = pathlib.Path(f"spiral_json/{pdb_id}.json") if f.exists(): data = json.loads(f.read_text()) lig = data["ligands"][0] hits.append({ "pdb": pdb_id, "cbqs": data["cbqs"], "C2endo": lig["ct_C2endo"], "bsa": lig["dSASA_contacted_residues"], "contacts": [c["idstr"] for c in lig["contacts"]], })

Explore

Browse by structure or RNA category

Search by PDB ID, ligand CCD code, or RNA category. Click any card to open in the 3D viewer with full interaction parameter annotations.

Reference

How to cite SPIRAL

If SPIRAL contributed to your research, please cite:

SPIRAL: Structural Pockets and Interacting RNA-Associated Ligands.
Lu X-J and Wang Y. (2026) Structural Pockets and Interacting RNA-Associated Ligands (SPIRAL): A DSSR-enabled Meta-Analysis of RNA-Small Molecule Recognition. bioRxiv. doi:10.64898/2026.05.19.726393 [preprint]

Lu XJ, Bussemaker HJ, Olson WK. (2015) DSSR: an integrated software tool for dissecting the spatial structure of RNA. Nucleic Acids Research 43(21):e142. doi:10.1093/nar/gkv716