RAPDORData

class RAPDOR.datastructures.RAPDORData(df: pandas.DataFrame, design: pandas.DataFrame, logbase: int = None, min_replicates: int = 2, control: str = 'Control', measure_type: str = 'Protein', measure: str = 'Intensities')

The RAPDORData Class storing results and containing functions for analysis

df

the dataframe that stores intensities and additional columns per protein.

Type:

pd.Dataframe

logbase

the logbase if intensities in df are log transformed. Else None.

Type:

int

design

dataframe containing information about the intensity columns in df

Type:

pd.Dataframe

array

(np.ndarray): The non-normalized intensities from the df intensity columns.

min_replicates

minimum number of replicates required to calculate scores

Type:

int

internal_design_matrix

dataframe where fraction columns are stored as a list instead of seperate columns

Type:

pd.Dataframe

norm_array

An array containing normalized values that add up to 1.

Type:

Union[None, np.ndarray]

distances

An array of size num_proteins x num_samples x num_samples that stores the distance between samples. If no distance was calculated it is None.

Type:

Union[None, np.ndarray]

permutation_sufficient_samples

Set to true if there are at least 5 samples per condition. Else False.

Type:

bool

score_columns

list of strings that are used as column names for scores that can be calculated via this object.

Type:

List[str]

control

Name of the level of treatment that should be used as control.

Type:

str

methods

List of supported distance functions

Type:

List[str]

Examples

An instance of the RAPDORData class is obtained via the following code. Make sure your csv files are correctly fomatted as desribed in the Data Prepatation Tutorial.

>>> df = pd.read_csv("../testData/testFile.tsv", sep="\t")
>>> design = pd.read_csv("../testData/testDesign.tsv", sep="\t")
>>> rapdor = RAPDORData(df, design, logbase=2)

__init__(df: pandas.DataFrame, design: pandas.DataFrame, logbase: int = None, min_replicates: int = 2, control: str = 'Control', measure_type: str = 'Protein', measure: str = 'Intensities')

Methods

__init__(df, design[, logbase, ...])
calc_all_anosim_value()	Calculates ANOSIM R for each protein and stores it in df
calc_all_permanova_f()	Calculates PERMANOVA F for each protein and stores it in df
calc_all_scores()	Calculates ANOSIM R, shift direction, peak positions and Mean Sample Distance.
calc_anosim_p_value(permutations, threads[, ...])	Calculates ANOSIM p-value via shuffling and stores it in df.
calc_distance_stats()	Calculates the mean distance and variance of this distance inside the same treatment group.
calc_distances([method])	Calculates between sample distances.
calc_distribution_features()	Calculates features used in a bubble plot
calc_mean_distance()	Calculates the distance between the means of the two treatment groups.
calc_permanova_p_value(permutations, threads)	Calculates PERMANOVA p-value via shuffling and stores it in df.
determine_peaks([beta])	Determines the Mean Distance, Peak Positions and shift direction.
determine_strongest_shift()	Determines the position of the strongest shift
export_csv(file[, sep])	Exports the extra_df to a file.
from_file(json_file)	Creates a class instance from a JSON file.
from_files(intensities, design[, logbase, sep])	Constructor to generate instance from files instead of pandas dataframes.
from_json(json_string)	Creates class instance from JSON string.
normalize_and_get_distances(method[, ...])	Normalizes the array and calculates sample distances.
normalize_array_with_kernel([kernel_size, eps])	Normalizes the array and sets norm_array attribute.
pca()	Performs PCA on the normalized array.
rank_table(values, ascending)	Ranks the df
remove_clusters()
run_preprocessing([method, kernel, impute, ...])	Normalizes the array, imputes missing values if needed and calculates sample distances.
to_json(file)	Exports the object to JSON
to_jsons()	encodes this object as a JSON string

Attributes

extra_df	Return a Dataframe Slice all columns from self.df that are not part of the intensity columns
methods
raw_lfc	Calculates the log2 fold change of the raw intensity means.
score_columns

calc_all_anosim_value()

Calculates ANOSIM R for each protein and stores it in df

calc_all_permanova_f()

Calculates PERMANOVA F for each protein and stores it in df

calc_all_scores()

Calculates ANOSIM R, shift direction, peak positions and Mean Sample Distance.

calc_anosim_p_value(permutations: int, threads: int, seed: int = 0, mode: str = 'local', callback=None)

Calculates ANOSIM p-value via shuffling and stores it in df. Adjusts for multiple testing.

Parameters:

• permutations (int) – number of permutations used to calculate p-value. Set to -1 to use all possible distinct permutations

• threads (int) – number of threads used for calculation

• seed (int) – seed for random permutation

• mode (str) – either local or global. Global uses distribution of R value of all proteins as background. Local uses protein specific distribution.

• callback (Callable) – A callback function that receives the progress in the form of a percent string e.g. “50”. This can be used in combination with a progress bar.

Returns:

fdr corrected p-values for each protein distribution (np.ndarray): distribution of R values used to calculate p-values

Return type:

p-values (np.ndarray)

calc_distance_stats()

Calculates the mean distance and variance of this distance inside the same treatment group.

calc_distances(method: str = None)

Calculates between sample distances.

Parameters:

method (str) – One of the values from methods. The method used for sample distance calculation.

Raises:

ValueError – If the method string is not supported or symmetric-kl-divergence is used without adding an epsilon to the protein intensities

calc_distribution_features()

Calculates features used in a bubble plot

Sets the features in current_embedding that can be used to plot a bubble plot of the data

calc_mean_distance()

Calculates the distance between the means of the two treatment groups.

calc_permanova_p_value(permutations: int, threads: int, seed: int = 0, mode: str = 'local')

Calculates PERMANOVA p-value via shuffling and stores it in df. Adjusts for multiple testing.

Parameters:

• permutations (int) – number of permutations used to calculate p-value

• threads (int) – number of threads used for calculation

• seed (int) – seed for random permutation

• mode (str) – either local or global. Global uses distribution of pseudo F value of all proteins as background. Local uses protein specific distribution.

Returns:

fdr corrected p-values for each protein distribution (np.ndarray): distribution of R values used to calculate p-values

Return type:

p-values (np.ndarray)

determine_peaks(beta: float = 1000)

Determines the Mean Distance, Peak Positions and shift direction.

The Peaks are determined the following way:

1. Calculate the mean of the norm_array per group (RNase & Control)

2. Calculate the mixture distribution of the mean distributions.

3. Calculate $D$ which is either:

   • Relative position-wise entropy of both groups to the mixture distribution if distance method is KL-Divergence or Jensen-Shannon

   • position-wise euclidean distance of both groups to the mixture distribution if distance method is Eucledian-Distance

4. Apply a soft-argmax to this using beta hyperparameter to find the relative position shift

determine_strongest_shift()

Determines the position of the strongest shift

export_csv(file: str, sep: str = ',')

Exports the extra_df to a file.

Parameters:

• file (str) – Path to file where dataframe should be exported to.

• sep (str) – seperator to use.

property extra_df

Return a Dataframe Slice all columns from self.df that are not part of the intensity columns

Returns:

slice of df

Return type:

pd.DataFrame

classmethod from_file(json_file)

Creates a class instance from a JSON file.

Returns:

the RAPDORData stored in the file.

Return type:

RAPDORData

classmethod from_files(intensities: str, design: str, logbase: int = None, sep: str = ',')

Constructor to generate instance from files instead of pandas dataframes.

Parameters:

• intensities (str) – Path to the intensities File

• design (str) – Path to the design file

• logbase (Union[None, int]) – Logbase if intensities in the intensity file are log transformed

• sep (str) – seperator used in the intensities and design files. Must be the same for both.

Returns: RAPDORData

classmethod from_json(json_string)

Creates class instance from JSON string.

Parameters:

json_string – string representation of the RAPDORData

Returns:

the RAPDORData stored in the string

Return type:

RAPDORData

normalize_and_get_distances(method: str, kernel: int = 0, eps: float = 0)

Normalizes the array and calculates sample distances.

Parameters:

• method (str) – One of the values from methods. The method used for sample distance calculation.

• kernel (int) – Averaging kernel size. This kernel is applied to the fractions.

• eps (float) – epsilon added to the intensities to overcome problems with zeros.

normalize_array_with_kernel(kernel_size: int = 0, eps: float = 0)

Normalizes the array and sets norm_array attribute.

Parameters:

• kernel_size (int) – Averaging kernel size. This kernel is applied to the fractions.

• eps (float) – epsilon added to the intensities to overcome problems with zeros.

pca()

Performs PCA on the normalized array.

Results are stored in df as PC1 and PC2. explained variance is stored in pca_var.

rank_table(values, ascending)

Ranks the df

This can be useful if you don´t have a sufficient number of samples and thus can`t calculate a p-value. The ranking scheme can be set via the function parameters.

Parameters:

• values (List[str]) – which columns to use for ranking

• ascending (List[bool]) – a boolean list indicating whether the column at the same index in values should be sorted ascending.

property raw_lfc

Calculates the log2 fold change of the raw intensity means.

Returns:

array containing the log2 fold change of the raw intensities

Return type:

np.ndarray

run_preprocessing(method: str = None, kernel: int = 0, impute: bool = False, impute_perc: float = 0.5, impute_nn: int = 10, impute_quantile: float = 0.95, eps: float = 0)

Normalizes the array, imputes missing values if needed and calculates sample distances.

Parameters:

• method (str) – One of the values from methods. The method used for sample distance calculation.

• kernel (int) – Averaging kernel size. This kernel is applied to the fractions.

• impute (bool) – Wheter to impute missing values using KNN

• impute_perc (float) – Maximum percentage of missing values per fraction and condition used for imputation.

• impute_nn (int) – N Neighbors to account in data imputation.

• impute_quantile (int) – Does only impute values for the quantile with the lowest mean distance between replicates. Others are assumed to be too noisy and data imputation might cause problems.

• eps (float) – epsilon added to the intensities to overcome problems with zeros.

to_json(file: str)

Exports the object to JSON

Parameters:

file (str) – Path to the file where the JSON encoded object should be stored.

to_jsons()

encodes this object as a JSON string

Returns:

JSON string representation of this object

Return type:

str