Design principles
- Table 3 first. The four sections of Table 3 — hardware conditions, assay conditions, construct parameters and factors in solution — are kept as the backbone, with the same units (nm, mW, pN/nm, K, kHz, Hz, mbar, mm/s).
- Technique-neutral core, technique-specific blocks. Fields shared by all force probes (detection, temperature, acquisition and filter rates, calibration) live at the top of
hardware; laser, magnet, cantilever, acoustic, centrifuge, nanopore, flow, BFP and tension-sensor parameters live in their own sub-blocks. - Identifiers over names. Every molecular entity carries resolvable identifiers (UniProt, PDB, GenBank, Rfam, ChEBI, Cell Ontology, NCBI Taxonomy, GO), with a
verifiedflag set when the ID was checked against its database. - Every value has evidence. Auto-populated values carry an
evidenceobject (method, snippet, confidence) so they can be audited and corrected. - Results are first-class. Reported numbers (unfolding forces, step sizes, Δx‡, lifetimes…) use a controlled observable vocabulary with canonical units, conditions and the model used, enabling cross-study meta-analysis.
- Data availability is explicit. Each record states whether a dataset is linked, where, how it was found, its file formats and whether it contains raw force traces.
Mapping to Table 3
How each Table 3 row maps to a schema path, and which experiment types it applies to (columns as in the original table; “all” = spans the full width).
Record structure
A record (smash-smfs-record.schema.json) describes one publication's experiment set. Top-level sections:
Techniques & force geometry
Supported techniques (technique.primary) and the force-application geometry vocabulary. Correlative modalities (confocal, TIRF, smFRET, IRM…) are recorded alongside.
Technique-specific hardware blocks
Experiment classes
Table 3's columns — nucleic acid or protein unfolding, DNA–protein interactions, force generation by cytoskeletal filaments or motors and rheology — are refined into sub-classes, and extended with receptor–ligand/adhesion bonds, membrane mechanics and methods.
Results schema
smash-smfs-results.schema.json defines a Result: one reported value (or range/distribution) for one observable on one entity or transition, under stated conditions, with the analysis model and provenance.
Observable vocabulary
Controlled names and canonical units. Values are stored in the canonical unit; the source unit is kept in the evidence snippet.
Raw trace format (TraceSet)
A technique-neutral JSON sidecar describing raw time series so optical, magnetic, acoustic, centrifugal, nanopore, AFM and flow data can be pooled. Containers can be HDF5 (including Lumicks .h5), TDMS, CSV or NumPy; the sidecar maps file paths to standard channel names, units and calibration, and optionally segments the protocol and annotates events.
Data availability status
Evidence & curation levels
Completeness score
fair_completeness reports the fraction of Table 3 fields populated in each section (hardware, assay, construct, solution, data) and their mean. It measures what is reported and recoverable, not experimental quality, and is a useful target for authors: a fully reported study scores 100%.
How records are built
- Harvest. Europe PMC is queried for each technique (optical/magnetic/torque/acoustic/centrifuge/nanopore tweezers, flow stretching, AFM-SMFS, BFP, micropipette, tension sensors, microneedles, plus generic SMFS phrases). DataCite is queried for deposits using the same vocabulary.
- Scope. A rule-based classifier keeps papers combining a force technique, a biological system and evidence of force applied between two ends (core), or SMFS instrument/analysis development (methods); cold-atom, colloid, plasmonic and similar physics papers are excluded.
- Link data. Datasets and code are linked through DataCite related identifiers, OpenAIRE ScholeXplorer, Europe PMC data-links, text-mined accessions and parsing of data-availability statements in open-access full text. Zenodo and Figshare file listings give formats and sizes.
- Annotate. Biological systems are matched with a curated gazetteer and resolved to UniProt (reviewed entries, organism-aware), Rfam, GenBank, ChEBI and Cell Ontology; organisms come from Europe PMC text mining. Hardware, assay and solution parameters are extracted from Methods text; results from abstracts.
- Curate. Hand-curated studies override automatic values and are marked
curated. Authors can submit verified records.
Downloads
All records (JSONL.gz)
Every record, schema v3, one JSON object per line.
Summary table (CSV)
One row per study: technique, system, identifiers, data status.
Deposits (CSV)
All repository deposits with linked studies and formats.
Record JSON Schema
Draft 2020-12, for validation.
Results & trace schema
Result, Observable, TraceSet definitions.
Simulation Corpus downloads
These files are the pulling-simulation corpus (SMASH-SIM v1.0.0). They are not experimental force-spectroscopy records. The schema and the field guide are on the simulation schema page.
Simulation records (JSONL.gz)
Every published simulation record, one JSON object per line.
Simulation summary (CSV)
One row per simulation: source, system, model, protocol, units and match level.
Simulation–experiment links (CSV)
Each link with its match level and the forces reported on each side.
SMASH-SIM record schema
SMASH-SIM v1.0.0. A simulation record, not an experimental SMFS record.
SMASH-SIM results schema
Simulation observables, including toughness, each with a unit system.