mol-hume

Molecular descriptors computed in C++, verified column by column against RDKit and Mordred.

mol-hume emits up to 1,269 descriptors per molecule in about 285 microseconds, plus a 2,048-bit ECFP alongside them. 1,109 of the descriptors reproduce definitions RDKit or Mordred already provide; 160 are new. Nothing is computed in Python.

pip install mol-hume

import molhume

X = molhume.featurize(["CCO", "CC(=O)Oc1ccccc1C(=O)O"], standardize="none")
# (2, 2670) float64: 622 minimal descriptors, then 2,048 ECFP bits

xgboost.XGBRegressor().fit(X, y)

featurize returns one array. Column names are identical for identical arguments, so they are available on request rather than returned every call:

df = pandas.DataFrame(X, columns=molhume.feature_names())

Pass both functions the same arguments and the names line up with the columns.

Selecting columns

One parameter, four ways to answer it:

molhume.featurize(smiles, columns="minimal")      # 622, the default
molhume.featurize(smiles, columns="full_no_new")  # 1,109 -- RDKit and Mordred definitions only
molhume.featurize(smiles, columns="full")         # all 1,269
molhume.featurize(smiles, columns=["TPSA", "AvgIpc", "BCUTc-1h"])   # these, in this order

column_set(name) returns the names in any of the three sets. ALL_COLUMNS lists every name a manual selection can use.

The selection decides what is computed, not only what is returned. A descriptor family none of whose columns are selected is not calculated, and neither are the individual eigensolves of the spectral family. Output is identical either way — tests/test_families.py compares every family and every set against an ungated run, cell for cell — but a narrow selection is cheaper as well as smaller. On 1,200 molecules of cpp/hard.smi, one thread, against 918 us/mol ungated:

selection	us/mol
`"minimal"` (622)	762	17% faster
`"full_no_new"` (1,109)	908	within noise
`"full"` (1,269)	900	within noise
`["TPSA", "ExactMolWt", "SLogP"]`	288	69% faster

The two full sets gain nothing: they request every family, so there is nothing to skip.

Sets and families

Column sets are the three choices above. In the paper's figures, where they appear beside other methods, they are written HUME_minimal, HUME_no_new and HUME_full; inside the package the prefix is redundant.

Families are the nineteen internal groupings the descriptors are computed in — autocorr, spectral, chi, estate, constit and so on. FAMILY_OFFSETS maps each to a half-open (start, stop) span:

lo, hi = molhume.FAMILY_OFFSETS["ringcount"]
ring_counts = X[:, lo:hi]                     # 47 columns, n5Ring .. nG12FAHRing

A set is what you request; a family is what gets computed. That is why minimal is cheaper — it needs none of autocorr, eta or pathcount — and full_no_new is not, since its 1,109 columns touch all nineteen.

qed

One column is in ALL_COLUMNS and in none of the three sets. qed costs 69.3 us/mol on its own (116 structural-alert subgraph searches, the most expensive column here) and is a drug-likeness score: a weighted geometric mean of eight properties already emitted as columns in their own right. full means every descriptor, not every expense, so it is opt-in:

molhume.featurize(smiles, columns=molhume.column_set("full", extra=["qed"]))
molhume.featurize(smiles, columns=["TPSA", "qed"])

OPTIONAL_COLUMNS names them. qed is appended after every other column, so opting in shifts nothing: column_set("full") is ALL_COLUMNS[:1269].

Standardization

standardize has no safe default. Descriptors are computed on the graph they are given, so a salt, a tautomer and a charge state are three different molecules and no library can infer which was meant. Leaving it unset warns once and lists the options.

value	effect
`"none"`	featurize exactly what was supplied
`"canonical"`	SMILES round-trip, nothing else
`"cleanup"`	RDKit `MolStandardize`: normalize, largest fragment, uncharge
a callable	your own `Mol -> Mol`

Passing "none" explicitly is silent; omitting it warns.

Arguments

argument	default	effect
`columns`	`"minimal"`	`"minimal"` (622), `"full_no_new"` (1,109), `"full"` (1,269), or a list of names in the order wanted. Decides what is computed as well as what is returned
`standardize`	`"none"`, warns if unset	what molecule the numbers describe
`threads`	`0`	descriptor-block workers; `0` is one per hardware thread. Pass `1` when the caller is already parallel; it costs about 3x
`fingerprint`	`True`	append `fp_size` ECFP bit columns after the descriptors, so descriptor indices do not shift when the flag changes. Off saves about 30 us/molecule that cannot be threaded
`fp_radius`	`3`	ECFP radius
`fp_size`	`2048`	ECFP bits
`on_error`	`"nan"`	a molecule that cannot be parsed or cannot be featurized: `"nan"` keeps the row and fills it, preserving alignment with the input; `"raise"` names the molecule and the reason; `"skip"` drops the row. Every molecule is isolated, so one failure never costs another its row
`dtype`	`float64`	`float32` halves memory and is what the boosting libraries convert to internally
`batch_size`	`4096`	rows per batch. Affects memory, not values

featurize also accepts RDKit Mol objects, which skips a parse.

import mol_hume returns the same module object. The distribution is mol-hume; import mol-hume is a syntax error, which no package can fix.

Verification

Every column was compared against its upstream definition over a 42,000-molecule corpus spanning 1 to 64 heavy atoms:

167 of 186 RDKit columns and 412 of 968 Mordred columns are bit-identical.
99.99% (RDKit) and 99.23% (Mordred) of values agree to within 1e-9.

The remainder are documented divergences rather than unexplained differences: cases where the upstream definition depends on atom numbering or on a Kekule choice and therefore has no single correct answer. Each is listed with a measurement in METHODS.md.

Timing

The 285 us figure is threaded, with threads=0 (one worker per hardware thread), the default. The descriptor block is the parallel part. Measured on a 12-thread M-series laptop over 4,000 corpus molecules:

	us/molecule
`threads=0` (default, 12 threads)	282
`threads=0`, `fingerprint=False`	247
`threads=1`	861
`threads=1`, `fingerprint=False`	846

threads=1 costs roughly 3x rather than 12x: the per-molecule boundary work does not parallelize. Use it when the caller is already parallel across processes.

RDKit version range

mol-hume requires rdkit>=2024.09.1,<2026.09, and this is a hard requirement. The library reads RDKit's MolPickler blob directly — a large part of where the speed comes from — and that format is explicitly not a stable API. Outside the measured range mol-hume refuses to import rather than misparse a molecule into wrong numbers with no symptom.

Within the range the format is checked rather than assumed. RDKit 2026.03 writes a different format version from 2025.09 and is accepted because 4,000 corpus molecules pickle to bytes differing only in the version triple, and all 1,269 columns over 8,000 molecules come out bit-identical. Widening the range for a future release is one command, tools/check_rdkit_release.py, plus work on the reader if the blobs did change. See MAINTENANCE.md.

The upper bound is loose deliberately: it is a courtesy to resolvers, stopping a fresh install from picking an RDKit years newer than anything measured, not a claim about 2027. If the pickle format does change, featurize raises an error naming the installed RDKit and what to do, the package still imports, and featurize_blocks(reader="api") still works on any RDKit, since it goes through the supported Python API.

Values are quoted against RDKit 2025.9.2 specifically. Perceived atom and bond properties drift across releases, so a different RDKit inside the range can still move the last digits.

Why 1,269 and not 1,539

The implemented set was 1,539 columns. Pairs carrying the same information were removed by a greedy cover in ascending compute cost: a column is dropped when a cheaper surviving column predicts it at |Spearman| >= 0.99 on ranks, and that must hold in every one of five heavy-atom strata rather than only on the pooled corpus — so a correlation that exists only because small and large molecules sit at opposite ends of both scales does not count. Columns that are NaN more than half the time, or that take one value for 99.9% of molecules, are dropped as unusable. 1,269 survive.

The minimal set

minimal-v2 is a 622-column subset, and the default since 0.7.0:

X = molhume.featurize(smiles)                     # the same call
X = molhume.featurize(smiles, columns="minimal")

It is a set, not a ranking. Every column was removed for one of three reasons, none of them a variance threshold:

the same physical quantity in different units — three electronegativity scales, atomic mass against atomic number, polarizability against volume. Read from the definitions, since no correlation cutoff separates "0.995, same construct" from "0.99, genuinely different";
already carried by the ECFP that ships alongside, or a duplicate of a count already emitted. Three fr_* flags go for the second reason: fr_halogen is [F,Cl,Br,I] against nF/nCl/nBr/nI/nX, fr_Ar_N is the SMARTS n, fr_bicyclic is [R2][R2];
an exact arithmetic identity of columns that remain — ring and constitutional counts that are sums of others, verified on two chemical spaces.

The expected descriptors are all present: molecular weight, Crippen logP, TPSA, H-bond donors and acceptors, rotatable bonds, ring counts, Kappa shape, chi connectivity, Labute ASA, Balaban J, Lipinski, and 72 of the 75 fr_* substructure flags.

The fr_* flags were dropped in 0.4.0 and restored in 0.5.0 and 0.6.0. They were dropped because they are detectable from the ECFP at AUROC 1.000, but that figure is conditional on the corpus rather than on the fingerprint. On a corpus with 5.4% salts the same measurement gives a median of 0.9929 and a floor of 0.786; fr_quatN reads 0.9995 on one corpus and 0.73 on the other. A 2x2 over radius (2 vs 3) and decoder (logistic vs XGBoost) moves the median by less than 0.001, so neither explains the gap. Detectability was never sound grounds for the drop. They are kept on mechanism: they encode curated assertions no structural descriptor derives — that a CYP enzyme attacks at a position, that a nitrogen is permanently charged, that a fragment is a toxicophore, which heteroatom sits in a ring, whether a hydroxyl is aliphatic or aromatic.

Measured cost

Benchmarked against the full 1,269 with the same untuned XGBoost head and the same 5-fold scaffold splits, on 29 of the 33 grid datasets:

panel	datasets	mean cost	worst
ADME and tox	10	-1.55%	+2.79%
physicochemical	6	-0.94%	+1.49%
classification	13	-0.17%	+1.99%
overall	29	-0.81%

Negative means the reduced set scored better. On none of the 29 datasets did the difference exceed that dataset's own fold-to-fold spread, and a sign test puts the reduced set behind on 11 of 29 (p = 0.27). The claim is no measurable difference at 49% of the columns, not that fewer columns help.

For contrast, the retired minimal-v1 cost +3.83% on the physicochemical panel at 800 columns. v2 is smaller and that loss is gone; the difference is what the two cut on. Full reasoning, decision by decision, is in HUME_Minimal_definition.md.

Platforms

Wheels are built for the platforms RDKit itself ships, since a mol-hume wheel for a platform with no RDKit wheel could not be imported:

	CPython 3.10 - 3.14
Linux x86_64, aarch64	manylinux_2_28
macOS arm64	11.0+
macOS x86_64	10.15+, needs `rdkit<=2025.9.2` (RDKit dropped Intel Mac after that)
Windows x86_64	MSVC

No musl, no 32-bit, no PyPy: RDKit publishes none of those. The extension links only the C++ runtime — no BLAS, no RDKit library, no NumPy ABI — so one wheel works across NumPy 1.x and 2.x.

Values are not bit-identical across architectures

This matters when comparing outputs between machines, and not at all when fitting a model. The exactness numbers above were measured on macOS arm64 with clang. The same source on x86-64 moves the last bits: 594 of the 1,269 columns under gcc, 595 under MSVC, with a maximum disagreement of 1.1e-14 of each column's range. Nothing structural changes; the NaN pattern is identical on all three.

No build flag removes this. The library reproduces upstream floating-point behavior, so a different libm's log and a different FMA decision are part of the result. CI measures it on every platform (tools/platform_drift.py) and the test suite asserts a bound, exactly rather than approximately on the reference platform.

Compare against each column's range rather than per value: several columns are differences that cancel to near zero (the centered autocorrelations, Cyclicity, DeltaMean), where a last-bit wobble reads as a relative error of 27.

Development

uv pip install -e . --python .venv/bin/python -c constraints.txt
.venv/bin/python -m pytest tests/

The pinned RDKit in constraints.txt is the oracle every exactness claim is measured against; install with -c constraints.txt or a bare editable install will silently upgrade it. tests/ runs in seconds against a committed fixture. The full exactness verifications against RDKit and Mordred are the root-level verify_*.py, which need the corpus and a second environment. See tests/README.md.

Acknowledgments

mol-hume reproduces descriptors first defined and published by two projects, and would not exist without either:

RDKit — Greg Landrum and contributors. RDKit parses the molecule and supplies every perceived atom and bond property this library computes from, and 186 of the emitted columns reproduce RDKit descriptor definitions. Several parameter tables here derive from published RDKit values, including the Crippen logP/MR atom-type contributions and the Hall-Kier alpha table. BSD 3-Clause.
Mordred — Hirotomo Moriwaki et al., J. Cheminform. 10, 4 (2018). 968 of the emitted columns reproduce Mordred definitions. BSD 3-Clause.

Where this library's values differ from either, the difference is deliberate and documented: cases where the upstream definition depends on atom numbering or on a Kekule choice and so has no single correct answer. Each is listed with a measurement in METHODS.md.

License

BSD 3-Clause. See LICENSE.

Development Status
- 4 - Beta
Intended Audience
- Science/Research
Programming Language
- C++
- Python :: 3
Topic
- Scientific/Engineering :: Chemistry

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mol_hume-0.9.2.tar.gz (506.4 kB view details)

Uploaded Sep 3, 2026 Source

Built Distributions

If you're not sure about the file name format, learn more about wheel file names.

The dropdown lists show the available interpreters, ABIs, and platforms. Enable javascript to be able to filter the list of wheel files.

mol_hume-0.9.2-cp314-cp314-win_amd64.whl (680.9 kB view details)

Uploaded Sep 3, 2026 CPython 3.14Windows x86-64

mol_hume-0.9.2-cp314-cp314-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (548.9 kB view details)

Uploaded Sep 3, 2026 CPython 3.14manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

mol_hume-0.9.2-cp314-cp314-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (510.2 kB view details)

Uploaded Sep 3, 2026 CPython 3.14manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

mol_hume-0.9.2-cp314-cp314-macosx_11_0_arm64.whl (475.3 kB view details)

Uploaded Sep 3, 2026 CPython 3.14macOS 11.0+ ARM64

mol_hume-0.9.2-cp314-cp314-macosx_10_15_x86_64.whl (522.1 kB view details)

Uploaded Sep 3, 2026 CPython 3.14macOS 10.15+ x86-64

mol_hume-0.9.2-cp313-cp313-win_amd64.whl (656.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.13Windows x86-64

mol_hume-0.9.2-cp313-cp313-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (548.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.13manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

mol_hume-0.9.2-cp313-cp313-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (509.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.13manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

mol_hume-0.9.2-cp313-cp313-macosx_11_0_arm64.whl (475.0 kB view details)

Uploaded Sep 3, 2026 CPython 3.13macOS 11.0+ ARM64

mol_hume-0.9.2-cp313-cp313-macosx_10_15_x86_64.whl (522.0 kB view details)

Uploaded Sep 3, 2026 CPython 3.13macOS 10.15+ x86-64

mol_hume-0.9.2-cp312-cp312-win_amd64.whl (656.5 kB view details)

Uploaded Sep 3, 2026 CPython 3.12Windows x86-64

mol_hume-0.9.2-cp312-cp312-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (548.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.12manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

mol_hume-0.9.2-cp312-cp312-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (509.4 kB view details)

Uploaded Sep 3, 2026 CPython 3.12manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

mol_hume-0.9.2-cp312-cp312-macosx_11_0_arm64.whl (475.0 kB view details)

Uploaded Sep 3, 2026 CPython 3.12macOS 11.0+ ARM64

mol_hume-0.9.2-cp312-cp312-macosx_10_15_x86_64.whl (521.9 kB view details)

Uploaded Sep 3, 2026 CPython 3.12macOS 10.15+ x86-64

mol_hume-0.9.2-cp311-cp311-win_amd64.whl (653.8 kB view details)

Uploaded Sep 3, 2026 CPython 3.11Windows x86-64

mol_hume-0.9.2-cp311-cp311-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (548.7 kB view details)

Uploaded Sep 3, 2026 CPython 3.11manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

mol_hume-0.9.2-cp311-cp311-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (510.0 kB view details)

Uploaded Sep 3, 2026 CPython 3.11manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

mol_hume-0.9.2-cp311-cp311-macosx_11_0_arm64.whl (473.5 kB view details)

Uploaded Sep 3, 2026 CPython 3.11macOS 11.0+ ARM64

mol_hume-0.9.2-cp311-cp311-macosx_10_15_x86_64.whl (519.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.11macOS 10.15+ x86-64

mol_hume-0.9.2-cp310-cp310-win_amd64.whl (653.3 kB view details)

Uploaded Sep 3, 2026 CPython 3.10Windows x86-64

mol_hume-0.9.2-cp310-cp310-manylinux_2_24_x86_64.manylinux_2_28_x86_64.whl (546.5 kB view details)

Uploaded Sep 3, 2026 CPython 3.10manylinux: glibc 2.24+ x86-64manylinux: glibc 2.28+ x86-64

mol_hume-0.9.2-cp310-cp310-manylinux_2_24_aarch64.manylinux_2_28_aarch64.whl (506.6 kB view details)

Uploaded Sep 3, 2026 CPython 3.10manylinux: glibc 2.24+ ARM64manylinux: glibc 2.28+ ARM64

mol_hume-0.9.2-cp310-cp310-macosx_11_0_arm64.whl (472.1 kB view details)

Uploaded Sep 3, 2026 CPython 3.10macOS 11.0+ ARM64

mol_hume-0.9.2-cp310-cp310-macosx_10_15_x86_64.whl (518.5 kB view details)

Uploaded Sep 3, 2026 CPython 3.10macOS 10.15+ x86-64

File details

Details for the file mol_hume-0.9.2.tar.gz.

File metadata

Download URL: mol_hume-0.9.2.tar.gz
Upload date: Sep 3, 2026
Size: 506.4 kB
Tags: Source
Uploaded using Trusted Publishing? Yes
Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mol_hume-0.9.2.tar.gz
Algorithm	Hash digest
SHA256	`46cac130d254a9a047c3964a4a1d1b2a386be29ca8d8836deb7245e0b8586561`
MD5	`50243ead090babe1010cb31cf97cec23`
BLAKE2b-256	`74ead63446c4446047c86812b8a39b92743570c8113bf20e7e86fee1549c279a`

Algorithm	Hash digest
SHA256	`3a6217b558c343bd6d0d8878ec73695809e5be0f5cfdbe59daaff4faa114dde5`
MD5	`ad4850050d92674647af009df0782615`
BLAKE2b-256	`acdb4094c8de3a191a6fb75da052287f67d8c98e531822722574fff3d8779ca9`

Algorithm	Hash digest
SHA256	`604340e1bf3728597b4d440e0f598691335ef69f580663acce4ee561465dd57c`
MD5	`ff2c1f3a121eaf30f2f896ed3e5a6602`
BLAKE2b-256	`9414fc30d996f9c2cd66a5ca5c993f008429b40688395df8320e817e58216617`

Algorithm	Hash digest
SHA256	`6efed9a0786e0267a1c1d7236e41c4454fd7b25e2a9f67cb91f6673926f874d7`
MD5	`5806847ca72e552d876da1a280004560`
BLAKE2b-256	`ccd5b0d82b4616492dbf236a9630ac06e52dc3015cd7f96bb742e43d45299490`

Algorithm	Hash digest
SHA256	`1a38a289692e4f3540f99433b56caae4e04912b55f0e5a082f4eda21c032e528`
MD5	`41b46c02fd6d27d9437f88405c2ffb4a`
BLAKE2b-256	`47bd7e700fda40e7cdd5ec48e114b0bc8692865939b403f1363fabb0cfb84b05`

Algorithm	Hash digest
SHA256	`01c350bd110768b0fcec0b2e1338f30b9bf89612ace9f0f6037193c3650a172a`
MD5	`f370c4f2c35daafb3a95274998341676`
BLAKE2b-256	`49cd15fcf3af9822d71ba505fa71144a4ac26d36de2551bbc53c92253d7b122d`

Algorithm	Hash digest
SHA256	`1f5850f1b3732dbb755e6d6de31002bace234607c960c5ac319cc22ca3ed5267`
MD5	`2aa20d33aec4e24c6bcecae0a8bb5d72`
BLAKE2b-256	`a655beb0ebb3cc2a5996a00013071a7566284557f8c77afbe5290ca39e8ae8f6`

Algorithm	Hash digest
SHA256	`d6bf50ece72a5df960f1ef341d3467ea32c0102e7ca8d9dce8fc7e8e3218c011`
MD5	`db509fdc11e7687e875b219d7ba92b09`
BLAKE2b-256	`c61ff0083f9d8c19a1b22e05830e8775111104de71d0f7fa5839f0d00a8b3a5b`

Algorithm	Hash digest
SHA256	`a989d6438bfa83ab8241d2f637a2a1be45ea7f72794700c8978321f81c121590`
MD5	`68f78e6319b150718a5e76e7beac51a4`
BLAKE2b-256	`8265bdd85a1394aefcdcdc422cda4fb9dbaca5ea060bc66f30fb4a737131bdf6`

Algorithm	Hash digest
SHA256	`df2fbc2fb6f7772e19e62815a1105aa96e0999711bcc23b7fdbb32d04a919bb2`
MD5	`6b1e49b2bd6b50120078fb34ce36caae`
BLAKE2b-256	`07bb8ed11078602ed6f02fc66a46886bef19641b95bf8784f85992cf3b1aa02e`

Algorithm	Hash digest
SHA256	`7e99967e975639e59799fd386db0cb57dbd98bbe5fa7e7ac743f44a6cc5e0a61`
MD5	`93ed6cec9d5dc61810022081505ab347`
BLAKE2b-256	`e3b63cd11bf9b0b2a295b206525841cabf48524d7fd114210e6a13893cbec0e9`

Algorithm	Hash digest
SHA256	`63b23cfbf6611f189173d3a9289d3eba6896925d8429348fabb8360168082c1c`
MD5	`c3ea9e727d2a44f60133126d97b7a7f6`
BLAKE2b-256	`dc7736488b1b16d3a560f1ea966f8a665fedcef86778a77faaf33d9aaf4f3d2e`

Algorithm	Hash digest
SHA256	`5395145882d465cbef662018124a6a51847c6b7ae156d223a60831a5fa99f75a`
MD5	`8bb1e75d34fe25c1e5d38d32b0a34a84`
BLAKE2b-256	`6c2282f486282fd6495e3e23efd190355479d2c8a9d06f7da8c3a713e4b35378`

Algorithm	Hash digest
SHA256	`41e4357e6d3c8fd81eebb44933bd83e4f2e98708e8f363b8e68e186702be1812`
MD5	`4efdedc783b99e47ad9c077ea3f1f2c5`
BLAKE2b-256	`474db9f83bcd6df8f92a777efd052e1966646e5828330f8e485eea5ea863cb8b`

Algorithm	Hash digest
SHA256	`1e80970a26efba22409ac9638ed91cdb6cff41dc6d15982b3aea0a3ed87f102f`
MD5	`7dfd0d8b6e3c20de527ac73564e7cc2b`
BLAKE2b-256	`01edb2cf25bb5e5b460674405e531877321452e62f116edca315c9058d42b063`

Algorithm	Hash digest
SHA256	`28af378e30e1c492e7f109e8d66ba4dc20f928b4f70b8bc666c2d2c55e349920`
MD5	`d0c2d2506dec491d57ed45c469d6e395`
BLAKE2b-256	`950b2b7b0bebc71958f09dedaedc126ba7325146f0c1ab86e738cd43d7eb8985`

Algorithm	Hash digest
SHA256	`5f61e882192fe43121d2f80ebb1770899b0b381539ef38731826c792e4068b9d`
MD5	`916e0308b0122ccb983276f261a347b9`
BLAKE2b-256	`ed4e32383f336363266bd9f47c0217b998814972ccf51ad9798201f0c9d42bbc`

Algorithm	Hash digest
SHA256	`e8f7844bb0be2e23d4d29fab19b9bad21a27a711d228eaa2b0d849697ae1d9a0`
MD5	`cce707c0dca16e118c1aab393050291d`
BLAKE2b-256	`c4d8caa2948cb3752c2a1a44c1a2c4d399b7ea95fa38806bd81f381419e6e199`

Algorithm	Hash digest
SHA256	`e025451819bbe11d5362b50be2335488010b8ac52423483484b774a84dae9e70`
MD5	`558818d8dc6851ab0b4524b6ab7b8829`
BLAKE2b-256	`1f52a251909c9de6e38fee25e4f8d9aa0a5cc5503a8bcc4d22a43edba32ba4e5`

Algorithm	Hash digest
SHA256	`6a61c3fc2a4523c86d5f64cdd627b122e484fedf8dcd0716e771bd6bebed13ef`
MD5	`aa28f9955f6366016c3147e19aabba57`
BLAKE2b-256	`fa000c9653991dcf05c652f8d934ab4faaccf65540bd70260d2442a9a854dcf7`

Algorithm	Hash digest
SHA256	`8a225ac83f95839efe9049aae76ee9a6adcb7a03fed4fc536232e74cf3a5c0ec`
MD5	`128b2b414babb5c5257af7ff95b792f8`
BLAKE2b-256	`972766e77da0c56b60a62abdf3312ea45f9a3366178a45340f569811ee92416a`

mol-hume 0.9.2

mol-hume

Selecting columns

Sets and families

qed

Standardization

Arguments

Verification

Timing

RDKit version range

Why 1,269 and not 1,539

The minimal set

Measured cost

Platforms

Values are not bit-identical across architectures

Development

Acknowledgments

License

Key dates

1 maintainer

Credits

License expression

Requires

Provides Extra

Tags

Classifiers

Download files

Source Distribution

Built Distributions

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Provenance

File details

File metadata

File hashes

Algorithm	Hash digest
SHA256	`069866abf4789b088728c93c6acaac7a1196b26b559a4c79d1ae5f2a3c28bdec`
MD5	`8c849e20791b226dd0032fbc92aa0ac1`
BLAKE2b-256	`4efa97e594bbdc8193dc661602886988072ca94d24d5418c19541b8d11ff33c7`

Algorithm	Hash digest
SHA256	`2f5589176a479c6019fbd9d24b11d6278b08eeb06d8fd77478aa944f2ab2fcbb`
MD5	`f7030e1305327ac18b0140faaee502a7`
BLAKE2b-256	`4ed1d6e537fe1f899b8e141b5453926063d81e5a9a51a0cdfd830ded28bc279c`

Algorithm	Hash digest
SHA256	`54d8c38da0f10bda1cd9aa8f1d9c3536cdfc1324c1ec75ca447f0a260efda8eb`
MD5	`6459c41306561a2b1b27fdc95f241638`
BLAKE2b-256	`cdaa90de7244003c0d9cf5fe98ea0f07daad2d5b7d30330f8d71e9f771edc7fc`

Algorithm	Hash digest
SHA256	`75182addc83f2672bbe868e668774c6655eaa9e677a52f59552e49d043f16ee5`
MD5	`e2d8e403ffa3f521ef2b789c85a28b85`
BLAKE2b-256	`f328c6dea5c479fc26dcf5a23c94bc4cf82aa93455b9f7712d1599b662bba394`

Algorithm	Hash digest
SHA256	`a15f74367df2765417a33088ee4ed453efa9b7a78b0722936f01e7d04ffcfee5`
MD5	`06293b6458849197c8c1ffe7d9cd644e`
BLAKE2b-256	`f88ac5b84130e36b847eb5861244a9a9dcbc25b3e5e17d9a14c88a266e4e7dcd`