helpers.pipeline

A dataclass that allows for an object oriented pipeline.

class lta.helpers.pipeline.Pipeline(file: pathlib.Path, output: pathlib.Path, n_rows_metadata: int, level: str, control: str, compartment: str, mode: str, sample_id: str, thresh: float, n: int)

The Lipid Traffic Analysis pipeline.

file

The path to the combined data input file.

Type: Path

output

Where to save the results.

Type: Path

n_rows_metadata

The number of rows to treat as column metadata. As Python is 0-indexed, passing 11 will read in rows 0-10.

Type: int

level

Metadata location of experimental conditions.

Type: str

control

Value within self.level that represents the control condition.

Type: str

compartment

Metadata location of sample tissue compartment.

Type: str

mode

Metadata location of lipidomics mode.

Type: str

sample_id

Metadata location of sample IDs.

Type: str

thresh

The fraction of samples that are 0 above which a lipid will be called 0 for a compartment.

Type: float

n

Number of bootstrap replicates.

Type: int

_calculate_enfc() → Dict[str, Dict[str, pandas.core.frame.DataFrame]]

Calculate error-normalised fold change.

Calculates the ENFC for each compartment across modes. For fold change to be meaningful, order must be specified. This will report fold-change as condition / self.control for all conditions except control within self.value.

Returns: Top level key is the experimental condition, mapped to a dictionary of modes and ENFC results
Return type: Dict[str, Dict[str, pd.DataFrame]]

_get_a_lipids() → Dict[str, pandas.core.frame.DataFrame]

Extract A-lipids from the dataset.

Any compartment where more than self.thresh of the samples are 0 is considered a total 0 for that lipid. Lipids that are non-0 for all compartments in any Phenotype are considered A-lipids.

Returns: Key is mode, value is table of A-lipids
Return type: Dict[str, pd.DataFrame]

_get_b_lipids(picky: bool = True) → Dict[str, pandas.core.frame.DataFrame]

Extract B-lipids from the dataset.

Any compartment where more than self.thresh of the samples are 0 is considered a total 0 for that lipid. Lipids that are non-0 for any pair of compartments within any Phenotype are considered B-lipids.

Notes

By definition, all A-lipids will also be B-lipids for every pair of compartments. These are referred to as B-consistent, or Bc for short. Those B-lipids that are not A-lipids are also known as B-picky, or Bp for short. Which set is calculated is controlled by the boolean flag picky.

Parameters: picky (bool) – default=True If true, do not consider lipids that are also A-lipids
Returns: Keys are the compartment pair and mode. Values are the table of B-lipids for that grouping.
Return type: Dict[str, pd.DataFrame]
Raises: AttributeError – If A-lipids have not been previously calculated

_get_n_lipids(n: int) → Dict[str, pandas.core.frame.DataFrame]

Extract N-lipids from the dataset.

Any compartment where more than self.thresh of the samples are 0 is considered a total 0 for that lipid. Lipids that are non-0 for n compartments in any Phenotype are considered N-lipids.

Notes

For historical consistency, N1-lipids (ie those found in only 1 compartment) are called U-lipids as they are Unique. Also N2-lipids are not the same as B-lipids. A B-lipid could occur in multiple pairs of compartment, while N2-lipids must only occur in 1.

Parameters: n (int) – The number of compartments to limit the search to
Returns: Keys are the compartment group and mode. Values are the table of N-lipids for that grouping.
Return type: Dict[str, pd.DataFrame]

_jaccard(data: Dict[str, pandas.core.frame.DataFrame], group: str) → Dict[str, Dict[str, pandas.core.frame.DataFrame]]

Calculate jaccard similarity and p-values.

This takes a dictionary of data. As the output of each group of lipids will be structured as such, it should be called per lipid group.

Notes

The P-values are calculated using a bootstrap approach on a centered Jaccard similarity.

Parameters

data (Dict[str, pd.DataFrame]) – A dictionary of modes and lipid data
group (str) – Which lipids are being checke. This is used by logging only.

Returns

Keys are the compartment group and mode. Values are the table of Jaccard similarity and p-values.

Return type

Dict[str, Dict[str, pd.DataFrame]]

run() → None

Run the full LTA pipeline.

This:

Calculates error-normalised fold change for all conditions relatve to control
Finds A-lipids and Jaccard distances.
Finds U-lipids and Jaccard distances.
Finds B-lipids (both picky and consistent) and Jaccard distances.
Finds N2-lipids and Jaccard distances.
Writes combined results.