Upload README.md with huggingface_hub

README.md

@@ -1,38 +1,41 @@
-# SPICE 1.1.4
+# SPICE (v1.1.4 `SPICE1` & v2.0.1 `SPICE2`)
 
 ## Overview
 SPICE (Small-molecule/Protein Interaction Chemical Energies) is a quantum-chemistry dataset
 of **single-point calculations** (ωB97M-D3(BJ)/def2-TZVPPD, Psi4) over many **conformations**
 of small molecules, dimers, ion pairs, and solvated amino acids / dipeptides, built for
-training machine-learned force fields. This is version **1.1.4**: **19,238 molecules /
-1,110,165 conformations** across **15 elements**. It is a **molecular** (non-periodic)
-**conformer-ensemble** dataset — stored the GEOM way, one conformer stack per molecule — and
-carries per-atom **forces** and per-conformer **energies**. A separate `SPICE2` loader wraps
-the larger v2.0.1 release (adds B, Si). SPICE 2.0 is *not* a superset of a stable v1; use v1.1.4
-to match the *Scientific Data* paper.
+training machine-learned force fields. It is a **molecular** (non-periodic) **conformer-ensemble**
+dataset — stored the GEOM way, one conformer stack per molecule — carrying per-atom **forces**
+and per-conformer **energies**. Two versions ship as two datasets with one shared loader:
+- **`SPICE1`** — v**1.1.4**: 19,238 molecules / 1,110,165 conformations, **15 elements**
+  (matches the *Scientific Data* paper).
+- **`SPICE2`** — v**2.0.1**: 113,985 molecules / 2,008,126 conformations, **17 elements**
+  (adds **B, Si**; new PubChem sets, an Amino-Acid–Ligand set, solvated PubChem, water clusters).
+
+SPICE 2.0 is *not* a strict superset of a stable v1; use `SPICE1` to reproduce the paper.
 
 ## Statistics
-<!-- from spice1_stats.json (atom/conformer ranges + element set exact; distributions from an 8,000-molecule sample) -->
-
-| property | value |
-|---|---|
-| molecules / conformations | 19,238 / 1,110,165 |
-| atoms per molecule (min / mean / max) | 2 / 33.4 / 96 |
-| conformers per molecule (min / mean / max) | 1 / 57.7 / 4,869 |
-| bounding-box diagonal, Å (sampled min / max) | 1.6 / 24.6 |
-| distinct elements | 15 |
-| element set | H, Li, C, N, O, F, Na, Mg, P, S, Cl, K, Ca, Br, I |
-| per-atom force magnitude, eV/Å (sampled min / mean / max) | 0.003 / 1.45 / 65.3 |
-| energy (`y`, total DFT energy, eV) | large & size-dependent (≈ −7×10⁵ … −1×10³) |
-| subset groups (`subset=`) | des370k (3,490), des_monomers (374), dipeptides (677), ion_pairs (28), pubchem (14,643), solvated_amino_acids (26) |
-
-Distributions in the figure are from a random sample of 8,000 molecules (one random conformer
-each); atom/conformer ranges and the element set are exact. `energy` is the **total** DFT energy,
-so its magnitude scales with molecule size (per-conformer *relative* energies are the useful signal).
+<!-- from spice1_stats.json / spice2_stats.json (atom/conformer ranges + element set exact; distributions from an 8,000-molecule sample) -->
+
+| property | SPICE1 (v1.1.4) | SPICE2 (v2.0.1) |
+|---|---|---|
+| molecules / conformations | 19,238 / 1,110,165 | 113,985 / 2,008,126 |
+| atoms per molecule (min / mean / max) | 2 / 33.4 / 96 | 2 / 49.5 / 110 |
+| conformers per molecule (min / mean / max) | 1 / 57.7 / 4,869 | 1 / 17.6 / 4,869 |
+| distinct elements | 15 | 17 |
+| element set | H, Li, C, N, O, F, Na, Mg, P, S, Cl, K, Ca, Br, I | + B, Si |
+| per-atom force magnitude, eV/Å (sampled min/mean/max) | 0.003 / 1.45 / 65.3 | 0.003 / 0.82 / 51.6 |
+| subset groups (`subset=`) | 6 (des370k, des_monomers, dipeptides, ion_pairs, pubchem, solvated_amino_acids) | 10 (+ amino_acid_ligand, pubchem_boron_silicon, solvated_pubchem, water_clusters) |
+
+Distributions in the figures (`spice1_histogram.png` / `spice2_histogram.png`) are from a random
+sample of 8,000 molecules (one random conformer each); atom/conformer ranges and the element set
+are exact. `energy` (`y`) is the **total** DFT energy, so its magnitude scales with molecule size
+(per-conformer *relative* energies are the useful signal). SPICE2 excludes 1 molecule that has zero
+conformations in the release.
 
 ## Usage / loading
 ```python
-from StructureCloud.Datasets import SPICE1
+from StructureCloud.Datasets import SPICE1, SPICE2         # SPICE2 has the identical interface
 from StructureCloud.Datasets import sample_by_relative, sample_by_boltzmann
 
 ds = SPICE1()                       # sample=True: one RANDOM conformer per molecule, eV units
@@ -46,6 +49,7 @@ SPICE1(sample_fn=sample_by_boltzmann(T=500))          # thermally-weighted confo
 SPICE1(subset='pubchem')                         # one subset group (see ds.available_subsets)
 SPICE1(units='raw')                              # native bohr / hartree / hartree·bohr⁻¹
 SPICE1(extra_fields=['mbis_charges'])            # attach extra fields (lazily downloaded)
+SPICE2(subset='pubchem_boron_silicon')           # SPICE2-only B/Si subset
 ```
 
 **Sampling** (one conformer returned per access): `sample=True` (default) random conformer per
@@ -76,7 +80,7 @@ atomic units):
 | `mayer_indices` | (N, N) | Mayer bond-order matrix | `_bonds` |
 | `wiberg_lowdin_indices` | (N, N) | Wiberg–Löwdin bond-order matrix | `_bonds` |
 
-The 4 `mbis_*` fields are **missing for 121 molecules** in v1.1.4 → NaN there.
+The 4 `mbis_*` fields are **missing for 121 molecules** (both versions) → NaN there.
 
 ## Preprocessed (collated) format
 Ships as a memory-mappable **main** file plus three **grouped, lazily-downloaded** extras files.
@@ -99,22 +103,30 @@ per-atom fields), `z_ptr` (one `z` per molecule), `conf_ptr` (per-conformer scal
 
 ## Splits
 One physical split `'all'` (one file per repo). Subsets are **not** splits — select one with the
-`subset=<id>` argument (`ds.available_subsets` = des370k, des_monomers, dipeptides, ion_pairs,
-pubchem, solvated_amino_acids). The `subset` attribute on each sample is the raw SPICE subset name.
+`subset=<id>` argument (`ds.available_subsets`). The `subset` attribute on each sample is the raw
+SPICE subset name. Groups: **SPICE1** — des370k, des_monomers, dipeptides, ion_pairs, pubchem,
+solvated_amino_acids (6). **SPICE2** — those + amino_acid_ligand, pubchem_boron_silicon,
+solvated_pubchem, water_clusters (10). PubChem "Set N" shards are merged into `pubchem`.
 
 ## Preprocessing & cleaning
 - All fields from the source HDF5 are preserved (common ones in the main file, the rest in the three
-  grouped extras files). Conformer stacks are kept per molecule (not flattened) so all three sampling
+  grouped extras files). Conformer stacks are kept per molecule (not flattened) so all sampling
   modes are available. `forces = −dft_total_gradient`.
 - Stored in raw units (bohr / hartree / hartree·bohr⁻¹); the loader converts to eV/Å by default.
   `energy`/`formation_energy` are stored float64 to preserve conformer-energy precision.
-- 121 molecules lack MBIS multipole fields in v1.1.4; their `mbis_*` extras are NaN.
+- **121 molecules lack the 4 `mbis_*` multipole fields** (both versions); their `mbis_*` extras are NaN.
+- **SPICE2 only:** 1 molecule with **zero conformations** in the release is dropped (113,986 → 113,985).
 
 ## Data distribution
-![atom / bounding-box / element / force / energy distributions](./spice1_histogram.png)
+SPICE1:
+![SPICE1 atom / bounding-box / element / force / energy distributions](./spice1_histogram.png)
+SPICE2:
+![SPICE2 atom / bounding-box / element / force / energy distributions](./spice2_histogram.png)
 
 ## Sources & citation
-- Data: SPICE 1.1.4, Zenodo DOI [10.5281/zenodo.8222043](https://doi.org/10.5281/zenodo.8222043)
-  (repo: <https://github.com/openmm/spice-dataset>). CC-BY-4.0.
+- SPICE 1.1.4: Zenodo DOI [10.5281/zenodo.8222043](https://doi.org/10.5281/zenodo.8222043).
+  SPICE 2.0.1: Zenodo DOI [10.5281/zenodo.10975225](https://doi.org/10.5281/zenodo.10975225).
+  Repo: <https://github.com/openmm/spice-dataset>. CC-BY-4.0.
 - Eastman et al., *SPICE, A Dataset of Drug-like Molecules and Peptides for Training Machine
   Learning Potentials*, Sci. Data 10, 11 (2023). DOI 10.1038/s41597-022-01882-6. arXiv:2209.10702.
+- Eastman et al., *Nutmeg and SPICE: Models and Data for Biomolecular Machine Learning* (SPICE 2), 2024.