A basic procedure for this pre-treatment of metabolomic data could follow
the three steps described thereafter:
i) removing of metabolites for which the proportion of missing data 
  (non detections) across all the samples is too high 
  (more than 20 to 50 percents according to your tolerance level);
ii) retrieving of missing values data using half minimum method 
  (i.e. half of the minimum value found for a metabolite across all samples);
iii) log-transformation of values.

If a scaling to the total intensity (normalization by sum of signals 
in each sample) or another normalization is necessary and pertinent, 
we recommend to do it before those three previously decribed steps.

For more information and options about pre-treatment of metabolomic data, 
see for example Liggi et al. 2018.

-- REFERENCES --

Liggi S, Hinz C, Hall Z, Santoru ML, Poddighe S, Fjeldsted J, ... & Griffin JL (2018).
KniMet: a pipeline for the processing of chromatography mass spectrometry 
metabolomics data. Metabolomics, 14(4), 52.