luna.interaction.fp.fingerprint module¶

Examples

>>> from luna.interaction.fp.fingerprint import CountFingerprint
>>> fp = CountFingerprint.from_bit_string("0010100110000010",
...                                       counts={2: 5, 4: 1, 7: 3, 8: 1, 14: 2})
>>> print(fp.indices)
[ 2  4  7  8 14]
>>> print(fp.counts)
{2: 5, 4: 1, 7: 3, 8: 1, 14: 2}

classmethod from_counts(counts, fp_length=4294967296, **kwargs)[source]¶

Initialize from a counting map.

Parameters

counts (dict) – Mapping between each index in indices to the number of counts.
fp_length (int) – The fingerprint length (total number of bits). The default value is \(2^{32}\).
**kwargs (dict, optional) – Extra arguments to CountFingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import CountFingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> on_bits = 8
>>> fp_length = 32
>>> counts = dict(zip(*np.unique(np.random.randint(0, fp_length, on_bits),
...                              return_counts=True)))
>>> print(counts)
{0: 1, 3: 2, 7: 1, 12: 1, 15: 1, 21: 1, 27: 1}
>>> fp = CountFingerprint.from_counts(counts=counts, fp_length=fp_length)
>>> print(fp.indices)
[ 0  3  7 12 15 21 27]
>>> print(fp.to_vector(compressed=False))
1 0 0 2 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0]

classmethod from_fingerprint(fp, **kwargs)[source]¶

Initialize from an existing fingerprint.

Parameters

fp (Fingerprint) – An existing fingerprint.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

classmethod from_indices(indices=None, counts=None, fp_length=4294967296, **kwargs)[source]¶

Initialize from an array of indices.

Parameters

indices (array_like of int, optional) – Indices of “on” bits. It is optional if counts is provided.
counts (dict, optional) – Mapping between each index in indices to the number of counts. If not provided, the default count value of 1 will be used instead.
fp_length (int) – The fingerprint length (total number of bits). The default value is \(2^{32}\).
**kwargs (dict, optional) – Extra arguments to CountFingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import CountFingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> on_bits = 8
>>> fp_length = 32
>>> indices, counts = np.unique(np.random.randint(0, fp_length, on_bits), return_counts=True)
>>> counts = dict(zip(indices, counts))
>>> print(counts)
{0: 1, 3: 2, 7: 1, 12: 1, 15: 1, 21: 1, 27: 1}
>>> fp = CountFingerprint.from_indices(indices, counts=counts, fp_length=fp_length)
>>> print(fp.indices)
[ 0  3  7 12 15 21 27]
>>> print(fp.to_vector(compressed=False))
[1 0 0 2 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0]

classmethod from_vector(vector, fp_length=None, **kwargs)[source]¶

Initialize from a vector.

Parameters

vector (numpy.ndarray or scipy.sparse.csr_matrix) – Array of counts.
fp_length (int, optional) – The fingerprint length (total number of bits). If not provided, the fingerprint length will be defined based on the vector shape.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import CountFingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> fp_length = 32
>>> vector = np.random.choice(5, size=(fp_length,), p=[0.76, 0.1, 0.1, 0.02, 0.02])
>>> print(vector)
[0 0 0 0 2 3 0 1 0 0 2 0 0 0 1 1 2 3 1 0 1 0 0 0 2 0 0 0 1 0 0 0]
>>> fp = CountFingerprint.from_vector(vector)
>>> print(fp.indices)
[ 4  5  7 10 14 15 16 17 18 20 24 28]
>>> print(fp.counts)
{4: 2, 5: 3, 7: 1, 10: 2, 14: 1, 15: 1, 16: 2, 17: 3, 18: 1, 20: 1, 24: 2, 28: 1}

get_count(index)[source]¶: Get the count value at index index. Return 0 if index is not in counts.

class Fingerprint(indices, fp_length=4294967296, unfolded_fp=None, unfolding_map=None, props=None)[source]¶

Bases: object

A fingerprint that stores indices of “on” bits.

Parameters

indices (array_like of int) – Indices of “on” bits.
fp_length (int) – The fingerprint length (total number of bits). The default value is \(2^{32}\).
unfolded_fp (Fingerprint or None) – The unfolded version of this fingerprint. If None, this fingerprint may have not been folded yet.
unfolding_map (dict, optional) – A mapping between current indices and indices from the unfolded version of this fingerprint what makes it possible to trace folded bits back to the original shells (features).
props (dict, optional) – Custom properties of the fingerprint, consisting of a string keyword and some value. It can be used, for instance, to save the ligand name and parameters used to generate shells (IFP features).

property bit_count¶

Number of “on” bits.

Type: int, read-only

calc_similarity(other)[source]¶

Calculates the Tanimoto similarity between this fingeprint and other.

Return type: float

Examples

>>> from luna.interaction.fp.fingerprint import Fingerprint
>>> fp1 = Fingerprint.from_bit_string("0010101110000010")
>>> fp2 = Fingerprint.from_bit_string("1010100110010010")
>>> print(fp1.calc_similarity(fp2))
0.625

property counts¶

Mapping between each index in indices to the number of counts, which is always 1 for bit fingerprints.

Type: dict, read-only

property density¶

Proportion of “on” bits in fingerprint.

Type: float, read-only

difference(other)[source]¶

Return indices in this fingerprint but not in other.

Return type

Raises

InvalidFingerprintType – If the informed fingerprint is not an instance of Fingerprint.
BitsValueError – If the fingerprints have different lengths.

fold(new_length=4096)[source]¶

Fold this fingerprint to size new_length.

Parameters: new_length (int) – Length of the new fingerprint, ideally multiple of 2. The default value is 4096.
Returns: Folded Fingerprint.
Return type: Fingerprint
Raises: BitsValueError – If the new fingerprint length is not a multiple of 2 or is greater than the existing fingerprint length.

Examples

>>> from luna.interaction.fp.fingerprint import Fingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> on_bits = 8
>>> fp_length = 32
>>> indices = np.random.randint(0, fp_length, on_bits)
>>> print(indices)
[12 15 21  0  3 27  3  7]
>>> fp = Fingerprint.from_indices(indices, fp_length=fp_length)
>>> print(fp.indices)
[ 0  3  7 12 15 21 27]
>>> print(fp.to_vector(compressed=False))
[1 0 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0]
>>> folded_fp = fp.fold(8)
>>> print(folded_fp.indices)
[0 3 4 5 7]
>>> print(folded_fp.to_vector(compressed=False))
[1 0 0 1 1 1 0 1]

property fp_length¶

The fingerprint length (total number of bits).

Type: int, read-only

classmethod from_bit_string(bit_string, fp_length=None, **kwargs)[source]¶

Initialize from a bit string (e.g. ‘0010100110’).

Parameters

bit_string (str) – String of 0s and 1s.
fp_length (int, optional) – The fingerprint length (total number of bits). If not provided, the fingerprint length will be defined based on the string length.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import Fingerprint
>>> fp = Fingerprint.from_bit_string("0010100110000010")
>>> print(fp.indices)
[ 2  4  7  8 14]
>>> print(fp.fp_length)
16

classmethod from_fingerprint(fp, **kwargs)[source]¶

Initialize from an existing fingerprint.

Parameters

fp (Fingerprint) – An existing fingerprint.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

classmethod from_indices(indices, fp_length=4294967296, **kwargs)[source]¶

Initialize from an array of indices.

Parameters

indices (array_like of int) – Indices of “on” bits.
fp_length (int) – The fingerprint length (total number of bits). The default value is \(2^{32}\).
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import Fingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> on_bits = 8
>>> fp_length = 32
>>> indices = np.random.randint(0, fp_length, on_bits)
>>> print(indices)
[12 15 21  0  3 27  3  7]
>>> fp = Fingerprint.from_indices(indices, fp_length=fp_length)
>>> print(fp.indices)
[ 0  3  7 12 15 21 27]
>>> print(fp.to_vector(compressed=False))
[1 0 0 1 0 0 0 1 0 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0]

classmethod from_rdkit(rdkit_fp, **kwargs)[source]¶

Initialize from an RDKit fingerprint.

Parameters

rdkit_fp (ExplicitBitVect or SparseBitVect) – An existing RDKit fingerprint.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

classmethod from_vector(vector, fp_length=None, **kwargs)[source]¶

Initialize from a vector.

Parameters

vector (numpy.ndarray or scipy.sparse.csr_matrix) – Array of bits.
fp_length (int, optional) – The fingerprint length (total number of bits). If not provided, the fingerprint length will be defined based on the vector shape.
**kwargs (dict, optional) – Extra arguments to Fingerprint. Refer to the documentation for a list of all possible arguments.

Return type

Examples

>>> from luna.interaction.fp.fingerprint import Fingerprint
>>> import numpy as np
>>> np.random.seed(0)
>>> fp_length = 32
>>> vector = np.random.choice([0, 1], size=(fp_length,), p=[0.8, 0.2])
>>> print(vector)
[0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 0 1 0 1 1 0 0 0 0 0 0 1 0 0 0 0]
>>> fp = Fingerprint.from_vector(vector)
>>> print(fp.indices)
[ 7  8 13 17 19 20 27]
>>> print(fp.fp_length)
32

get_bit(index)[source]¶

Get the bit/count value at index index.

Raises: BitsValueError – If the provided index is in a different bit scale.

get_num_bits()[source]¶: Get the fingerprint length (total number of bits).

get_num_off_bits()[source]¶: Get the number of “off” bits.

get_num_on_bits()[source]¶: Get the number of “on” bits.

get_on_bits()[source]¶

Get “on” bits.

Return type: numpy.ndarray

get_prop(key)[source]¶: Get value of the property key. If not set, raise KeyError.

property indices¶

Indices of “on” bits.

Type: array_like of int, read-only

intersection(other)[source]¶

Return the intersection between indices of two fingerprints.

Return type

Raises

InvalidFingerprintType – If the informed fingerprint is not an instance of Fingerprint.
BitsValueError – If the fingerprints have different lengths.

property name¶

The property ‘name’. If it was not provided, then return an empty string.

Type: str

property num_levels¶

The property ‘num_levels’ used to generate this fingerprint (see ShellGenerator). If it was not provided, then return None.

Type: int

property num_shells¶

The property ‘num_shells’ (see ShellGenerator). If it was not provided, then return None.

Type: int

property props¶

The custom properties of the fingerprint.

Type: dict, read-only

property radius_step¶

The property ‘radius_step’ used to generate this fingerprint (see ShellGenerator). If it was not provided, then return None.

Type: float

set_prop(key, value)[source]¶: Set value to the property key.

symmetric_difference(other)[source]¶

Return indices in either this fingerprint or other but not both.

Return type

numpy.ndarray or scipy.sparse.csr_matrix

Raises

InvalidFingerprintType – If the informed fingerprint is not an instance of Fingerprint.
BitsValueError – If the fingerprints have different lengths.

to_bit_string()[source]¶

Convert this fingerprint to a string of bits.

Warning

This function may raise a MemoryError exception when using huge indices vectors. If you found this issue, you may want to try a different data type or apply a folding operation before calling to_bit_string.

Return type: str
Raises: MemoryError – If the operation ran out of memory.

to_bit_vector(compressed=True)[source]¶

Convert this fingerprint to a vector of bits.

Warning

This function may raise a MemoryError exception when using huge indices vectors. If you found this issue, you may want to try a different data type or apply a folding operation before calling to_bit_vector.

Parameters

compressed (bool) – If True, build a compressed sparse matrix (scipy.sparse.csr_matrix).

Returns

Vector of bits/counts. Return a compressed sparse matrix (scipy.sparse.csr_matrix) if compressed is True. Otherwise, return a Numpy array (numpy.ndarray)

Return type

Raises

BitsValueError – If some of the fingerprint indices are greater than the fingerprint length.
MemoryError – If the operation ran out of memory.

to_rdkit(rdkit_fp_cls=None)[source]¶

Convert this fingerprint to an RDKit fingerprint.

Note

If the fingerprint length exceeds the maximum RDKit fingerprint length (\(2^{31} - 1\)), this fingerprint will be folded to length \(2^{31} - 1\) before conversion.

Returns: If fp_length is less than \(1e5\), ExplicitBitVect is used. Otherwise, SparseBitVect is used.
Return type: ExplicitBitVect or SparseBitVect

to_vector(compressed=True, dtype=<class 'numpy.int32'>)[source]¶

Convert this fingerprint to a vector of bits/counts.

Warning

This function may raise a MemoryError exception when using huge indices vectors. If you found this issue, you may want to try a different data type or apply a folding operation before calling to_vector.

Parameters

compressed (bool) – If True, build a compressed sparse matrix (scipy.sparse.csr_matrix).
dtype (data-type) – The default value is np.int32.

Returns

Vector of bits/counts. Return a compressed sparse matrix (scipy.sparse.csr_matrix) if compressed is True. Otherwise, return a Numpy array (numpy.ndarray)

Return type

numpy.ndarray or scipy.sparse.csr_matrix

Raises

BitsValueError – If some of the fingerprint indices are greater than the fingerprint length.
MemoryError – If the operation ran out of memory.

unfold()[source]¶

Unfold this fingerprint and return its parent fingerprint.

Return type: Fingerprint

property unfolded_fp¶

The unfolded version of this fingerprint. If None, this fingerprint may have not been folded yet.

Type: Fingerprint or None, read-only

property unfolded_indices¶

Indices of “on” bits in the unfolded fingerprint.

Type: array_like of int, read-only

property unfolding_map¶

The mapping between current indices and indices from the unfolded version of this fingerprint what makes it possible to trace folded bits back to the original shells (features).

Type: dict, read-only

union(other)[source]¶

Return the union of indices of two fingerprints.

Return type