Skip to contents

Train a new FastRet model (FRM) for retention time prediction

Source: R/train.R
train_frm.Rd

Trains a new model from molecule SMILES to predict retention times (RT) using the specified method.

Usage

train_frm(
  df,
  method = "lasso",
  verbose = 1,
  nfolds = 5,
  nw = 1,
  degree_polynomial = 1,
  interaction_terms = FALSE,
  rm_near_zero_var = TRUE,
  rm_na = TRUE,
  rm_ns = FALSE,
  seed = NULL,
  do_cv = TRUE
)

Arguments

df

A dataframe with columns "NAME", "RT", "SMILES" and optionally a set of chemical descriptors. If no chemical descriptors are provided, they are calculated using the function preprocess_data().

method

A string representing the prediction algorithm. Either "lasso", "ridge", "gbtree", "gbtreeDefault" or "gbtreeRP". Method "gbtree" is an alias for "gbtreeDefault".

verbose

A logical value indicating whether to print progress messages.

nfolds

An integer representing the number of folds for cross validation.

nw

An integer representing the number of workers for parallel processing.

degree_polynomial

An integer representing the degree of the polynomial. Polynomials up to the specified degree are included in the model.

interaction_terms

A logical value indicating whether to include interaction terms in the model.

rm_near_zero_var

A logical value indicating whether to remove near zero variance predictors.

rm_na

A logical value indicating whether to remove NA values before training. Highly recommended to avoid issues during model fitting. Setting this to FALSE with method = "lasso" will most likely lead to errors.

rm_ns

A logical value indicating whether to remove chemical descriptors that were considered as not suitable for linear regression based on a previous analysis of an independent dataset. Currently not used.

seed

An integer value to set the seed for random number generation to allow for reproducible results.

do_cv

A logical value indicating whether to perform cross-validation. If FALSE, the cv element in the returned object will be NULL.

Value

A 'FastRet Model', i.e., an object of class frm. Components are:

  • model: The fitted base model. This can be an object of class glmnet (for Lasso or Ridge regression) or xgb.Booster (for GBTree models).

  • df: The data frame used for training the model. The data frame contains all user-provided columns (including mandatory columns RT, SMILES and NAME) as well the calculated chemical descriptors. (But no interaction terms or polynomial features, as these can be recreated within a few milliseconds).

  • cv: A named list containing the cross validation results, or NULL if do_cv = FALSE. When not NULL, elements are:

    • folds: A list of integer vectors specifying the samples in each fold.

    • models: A list of models trained on each fold.

    • stats: A list of vectors with RMSE, Rsquared, MAE, pBelow1Min per fold.

    • preds: Retention time predictions obtained in CV as numeric vector.

  • seed: The seed used for random number generation.

  • version: The version of the FastRet package used to train the model.

  • args: The value of function arguments besides df as named list.

Examples

m <- train_frm(df = RP[1:40, ], method = "lasso", nfolds = 2, verbose = 0)
# For the sake of a short runtime, only the first 40 rows of the RP dataset
# are used in this example. In practice, you should always use the entire
# training dataset for model training.