This research work deals with the use of a previously published algorithm to filter out redundant spectral data from MALDI-TOF MS measurements with intact bacteria (“dereplication” algorithm). This approach was found applicable in culturomics studies. Such studies are considered complementary to metagenomics for the description of bacterial diversity. The so-called SPeDE algorithm reduces the datasets to genomically diverse operational isolation units (OIUs). It compares spectra and recognizes differences based on unique spectral features (USFs). Both hierarchical clustering and network analyses were applied here and compared to assess the similarity of isolates and the related references based on the USF matrix generated by SPeDE. The algoritm itself does not visualize such relationships. The distances in the hierarchical clustering dendrogram were based on the number of USFs between the references of each OIU. Conversely, the network analysis visualizes all comparisons for which no USFs were detected. First, previous datasets for Lactobacillus and Burkholderia were processed, which allowed finding out that the clustering option provides less confidential assessments as it often could not cluster correctly more references belonging to the same isolate. On the other hand, the network analysis was found much more convincing in this regard. In a case study, a soil sample was cultivated, divided, subcultivated and further processed by a pipetting robot to make MALDI probes. Then MALDI-TOF MS spectral profiles were acquired. All datasets were processed by SPeDE and the USF matrix utilized to construct a dendrogram or a network graphics. The number of spectra attributed to the genus Burkholderia was 200 providing 43 OIUs matching mostly to 16 references. The dendrogram failed in distinguishing several clusters of references that were apparent in the network. Fourteen selected isolates were additionally subjected to an Illumina-platform DNA sequencing to get further clues for evaluation, which finally confirmed that the network visualization based on USFs in MALDI-TOF mass spectra enabled a superior selection of genomically diverse OIUs compared to the hierarchical clustering.

This is a very interesting and well written manuscript providing a novel insight into MALDI-TOF mass spectrometry of intact bacteria and related bioinformatics. The obtained results are well evaluated and relevantly discussed. The benefit for the community of readers-researchers is obvious as the use of the described strategy offers a more reliable species differentiation and reading of biological information from MALDI-based spectral profiles of bacterial cells.

A minor comment: Lines 207 and 282/286 – discrepancy in the number of OUIs, 43 vs. 44