Machine Learning in Mass Spectrometry: A MALDI-TOF MS Approach to Phenotypic Antibacterial Screening.

Machine learning techniques can be applied to MALDI-TOF mass spectral data of drug-treated cells to obtain classification models which assign the mechanism of action of drugs. Here, we present an example application of this concept to the screening of antibacterial drugs that act at the major bacterial target sites such as the ribosome, penicillin-binding proteins, and topoisomerases in a pharmacologically relevant phenotypic setting. We show that antibacterial effects can be identified and classified in a label-free, high-throughput manner using wild-type Escherichia coli and Staphylococcus aureus cells at variable levels of target engagement. This phenotypic approach, which combines mass spectrometry and machine learning, therefore denoted as PhenoMS-ML, may prove useful for the identification and development of novel antibacterial compounds and other pharmacological agents.

High-Accuracy Mass Spectrometry Based Screening Method for the Discovery of Cysteine Containing Peptides in Animal Venoms and Toxins.

Venom and toxin samples derived from animal origins are a rich source of bioactive peptides. A high proportion of bioactive peptides that have been identified in venom contain one or more disulfide bridges, which are thought to stabilize tertiary structure, and therefore influence the peptides' specificity and activity. In this chapter, we describe a label-free mass spectrometry-based screening workflow specifically to detect peptides that contain inter- and intramolecular disulfide bonds, followed by elucidation of their primary structure. This method is based on the determination of the normalized isotope shift (NIS) and the normalized mass defect (NMD) of peptides, two parameters which are heavily influenced by the presence of sulfur in a peptide, where cysteines are the main contributing residues. Using ant defensive secretions as an example, we describe the initial fractionation of the venom on strong cation exchange followed by nanoflow HPLC and mass spectrometry. High resolution zoom scan spectra of high-abundance peptides are acquired, allowing an accurate determination of both monoisotopic and average mass, which are essential for calculation of NMD and NIS. Candidate peptides exhibiting relative low NMD and high NIS values are selected for targeted de novo sequencing. By fine-tuning the collision energy for optimal fragmentation of each selected precursor ions, the full sequence of several novel inter- and intramolecular disulfide bond containing ant defensive peptides can be established.

Screening Method for the Discovery of Potential Bioactive Cysteine-Containing Peptides Using 3D Mass Mapping.

Animal venoms and toxins are a valuable source of bioactive peptides with pharmacologic relevance as potential drug leads. A large subset of biologically active peptides discovered up till now contain disulfide bridges that enhance stability and activity. To discover new members of this class of peptides, we developed a workflow screening specifically for those peptides that contain inter- and intra-molecular disulfide bonds by means of three-dimensional (3D) mass mapping. Two intrinsic properties of the sulfur atom, (1) its relatively large negative mass defect, and (2) its isotopic composition, allow for differentiation between cysteine-containing peptides and peptides lacking sulfur. High sulfur content in a peptide decreases the normalized nominal mass defect (NMD) and increases the normalized isotopic shift (NIS). Hence in a 3D plot of mass, NIS, and NMD, peptides with sulfur appear in this plot with a distinct spatial localization compared with peptides that lack sulfur. In this study we investigated the skin secretion of two frog species; Odorrana schmackeri and Bombina variegata. Peptides from the crude skin secretions were separated by nanoflow LC, and of all eluting peptides high resolution zoom scans were acquired in order to accurately determine both monoisotopic mass and average mass. Both the NMD and the NIS were calculated from the experimental data using an in-house developed MATLAB script. Candidate peptides exhibiting a low NMD and high NIS values were selected for targeted de novo sequencing, and this resulted in the identification of several novel inter- and intra-molecular disulfide bond containing peptides. Graphical Abstract ᅟ.