k_seq.model: kinetics and pool models

class k_seq.model.ModelBase(**params)

Bases: abc.ABC

base structure for a model A model contains following aspects:

  • model param estimation: need data, _get_mask, params, seed,

  • random variable generator, need params, seed, _get_mask

abstract func(**params)

model function: params –> prediction

abstract predict(**params)

Function to predict model output

k_seq.model.count module

Model the sequencing process to generate counts

k_seq.model.count.multinomial(p, N, seed=None)

Multinomial distribution for a given probability p and total number of draws

k_seq.model.kinetic module

A collection of commonly used kinetic models

Notes

kinetic models were currently implemented as callable function. It might migrate to Model subclass for

storing parameters, set required parameters, etc

TODO: fix broadcast problem: including broadcast as argument changed the signature for the model function

class k_seq.model.kinetic.BYOModel

Bases: object

A collection of BYO kinetic models where some values are fixed:

exp time (t): 90 min BYO degradation factor ($lpha$): 0.479

models:

  • reacted_frac(broadcast=False)

  • amount_first_order(broadcast=False)

  • composition_first_order(c, p0, k, A)

Parameters

  • p0 (1-D list-like) – Amount or composition of sequences in the input pool

  • c (float or 1-D list-like) – Concentration of substrates, return input pool if negative

  • k (float or 1-D list-like) – Kinetic coefficient for each sequence in first-order kinetics, with length as number of sequences

  • A (float or 1-D list-like) – Asymptotic conversion for each sequence in first-order kinetics, with length as number of sequences

return: concentration/composition at the name (time, concentration) point

static amount_first_order(broadcast=False)

Get model for absolute amount of reacted pool, if c is negative, return 0 Notice: broadcast should be False if multiple p0 is used.

Parameters

broadcast (bool) – if True, apply broadcast for k and A, outer product between A and k are calculated;if False, k and A have to be same length and apply elementwise production. Default False.

Returns

p0 (1-D list-like): Amount or composition of sequences in the input pool

c (float or 1-D list-like): Concentration of substrates, return input pool if negative k (float or 1-D list-like): Kinetic coefficient for each sequence in first-order kinetics, with length as number of sequences A (float or 1-D list-like): Asymptotic conversion for each sequence in first-order kinetics, with length as number of sequences

which returns:

Absolute amount of each sequence in each sample float, 1-D or 2-D np.ndarray with shape (uniq_seq_num, sample_num)

Return type

a callable with arguments

static composition_first_order(c, p0, k, A)

Function of pool composition w.r.t. BYO concentration (c) if x < 0, output is p0

broadcast is not supported that is k and A should have same length

Parameters

  • c (float or 1-D list-like) – Concentration of substrates, return input pool if negative

  • p0 (1-D list-like) – Amount or composition of sequences in the input pool

  • k (float or 1-D list-like) – Kinetic coefficient for each sequence in first-order kinetics, with length as number of sequences

  • A (float or 1-D list-like) – Asymptotic conversion for each sequence in first-order kinetics, with length as number of sequences

Returns

Pool composition for sequences in each sample float, 1-D or 2-D np.ndarray with shape (len(k), sample_num)

static reacted_frac(broadcast=False)

Get model for Reacted fraction for each seq in a pool :param broadcast: if True, apply broadcast for k and A, outer product between A and k are calculated;if False, k and A have to be same length and apply elementwise production. Default False. :type broadcast: bool

Returns

c (float or 1-D list-like): Concentration of substrates, return input pool if negative

k (float or 1-D list-like): Kinetic coefficient for each sequence in first-order kinetics, with length as number of sequences A (float or 1-D list-like): Asymptotic conversion for each sequence in first-order kinetics, with length as number of sequences

which returns:

Reacted fraction for each sequence in each sample float, 1-D or 2-D np.ndarray with shape (uniq_seq_num, sample_num)

Return type

a callable with arguments

k_seq.model.kinetic.check_scalar(value)

Check if value is a scalar or is the single value in a vector/matrix/tensor

k_seq.model.kinetic.first_order(c, k, A, alpha, t, broadcast=False)

Base first-order kinetic model, returns reacted fraction of input seqs given parameters

broadcast are available on A, k, c and a full return tensor will have shape (A, k, c) if any of these 3 parameters is scalar, the dimension is automatically squeezed while maintaining the order

Note: for c_i < 0, returns ones as it is input pool

Args: c (float or 1-D list-like): Concentration of substrates, return input pool if negative

k (float or 1-D list-like): Kinetic coefficient for each sequence in first-order kinetics, with length as number of sequences A (float or 1-D list-like): Asymptotic conversion for each sequence in first-order kinetics, with length as number of sequences alpha (float): degradation coefficient for substrates t (float): Reaction time in the experiments broadcast (bool): if True, apply broadcast for k and A, outer product between A and k are calculated;if False, k and A have to be same length and apply elementwise production. Default False.

k_seq.model.kinetic.to_scalar(value)

Try to convert a value to scalar, if possible

k_seq.model.pool module

Model for a sequence pool with constituent change from reaction

class k_seq.model.pool.PoolModel(count_model, kinetic_model=None, param_table=None, note=None, **params)

Bases: k_seq.model.ModelBase

Model of a kinetic pool,

kinetic_model

input initial pool with parameter and return a reacted pool composition

Type

callable

count_model

input pool composition and return a list of counts given total counts or params

Type

callable

kinetic_params

list of parameter names for kinetic model

Type

list of str

count_params

list of parameter names for count model

Type

list of str

note

note for the model

Type

str

__init__(count_model, kinetic_model=None, param_table=None, note=None, **params)

Initialize a pool model with given kinetic models and count_model :param count_model: model for sequencing counts :type count_model: ModelBase or callable :param kinetic_model: model for pool kinetics, no react if not given :type kinetic_model: ModalBase or callable :param **params:

func(**params)

Draw counts from given parameters

Returns

reacted amount if not normalized counts for each sequence

Return type

output sum from kinetic model

predict(**params)

Wrapper over _get_mask, can accept parameters to overwrite current ones if exist