Experts in bioinformatics find 79 candidates for new antibiotics in human faeces

When we talk about antibiotics, we primarily think about medicinal drugs that can be bought at the chemist. However, antibiotics are primarily microscopic bullets by which bacteria kill each other. People have simply taken and adapted part of these biologically active substances for their needs – the fight against infections. Since various microorganisms have the ability to produce such substances, there is a question: can the bacteria that live in our guts (of which there are up to a thousand different species) produce antibiotics?

To find an answer to this question, an international team of scientists has created a new computational method for searching for cyclopeptides. The team is headed by Pavel Pevzner, Chief Advisor of the Center for Algorithmic Biotechnology at St Petersburg University and professor at the University of California at San Diego (UCSD), and one of the leading experts in bioinformatics. They are a pharmacologically important class of substances that includes a lot of well-known antibiotics, antitumor compounds, and immunosuppressive drugs. On the one hand, the cyclic structure of such molecules gives them important physicochemical properties that make it possible to use them as drugs. On the other hand, it significantly complicates their detection in natural samples studied by scientists.

Gramicidin S is a typical representative of cyclic peptides. It consists of only five different amino acids, each of which is repeated twice in a cycle.

Gramicidin S was discovered by Soviet scientists Georgy Gause and Maria Brazhnikova in 1942. Since then, a total of about 1,200 cyclic peptides have been found and described. Work on discovering each new compound takes many years due to the difficulties in detecting such substances and determining their structure. Breakthroughs in recent years in biotechnology and bioinformatics have brought us closer to the emergence of a much faster way to find previously unknown cyclic peptides. An article by Pevzner’s team, recently published in the respected journal Cell Systems, describes more than 400 new cyclic peptides at once. To identify them, it took the team of scientists only a few days as they were using their new algorithmic approach, called CycloNovo, on a computing cluster. During this time, they processed a large amount of experimental data from samples of microbes and plants obtained earlier by other scientists. They had published them in the public domain in the hope that appropriate processing algorithms might appear.

CycloNovo analyses the output of modern highly precise mass spectrometers. They are extremely sensitive ‘molecular weights’ that break down molecules into small fragments and measure their mass. By analysing the masses of molecular fragments, CycloNovo determines which mass spectra correspond to cyclic peptides rather than standard linear compounds. Then, this algorithm reveals their structure – the amino acid sequence. The whole procedure can be compared to finding a needle in a haystack. This computational problem has faced science for more than ten years, and it has only now become possible to completely solve it. It is curious that the proposed algorithm resembles the approach used to solve the genome assembly problem. This is a completely different field of bioinformatics, to which the Center for Algorithmic Biotechnology has also made a significant contribution.

The very first applications of CycloNovo to real data demonstrated its applied significance as a tool in the hands of biologists. The applications have already resulted in interesting results from a medical point of view. So, with the help of the programme, the mass spectra of the obtained samples of human faeces were analysed. They had been collected for four years by Professor Larry Smarr, one of the co-authors of the article. In them, a large number of cyclopeptides characteristic of flax seeds were unexpectedly found, which are not bacteria that are part of the human microbiota. At first glance, this fact calls into question the correctness of the algorithm and the study as a whole. However, it soon became clear that Smarr consumed linseed oil as part of his diet, and the days of use exactly coincide with the dates at which the corresponding samples were collected. Thus, for the first time, it was clearly demonstrated that cyclopeptides from food can maintain their structure when passing through the unfavourable environment of the human stomach and intestines. It is likely that cyclopeptides are the main factor explaining the antimicrobial activity of flax seeds. Another important result of this experiment was that flax seed cyclopeptides are just a few representatives from the 79 cyclopeptides found by CycloNovo in the faeces samples under study. Determining their biological significance and functions is a separate task, on which work is already underway in collaboration with physicians and biologists. It is planned that various approaches to the processing of additional types of data obtained from the human microbiome will be added to the existing analysis.

Alongside Pavel Pevzner, young scientists from St Petersburg University, Andrey Prjibelski and Alexey Gurevich, took part in the research.  

An accurate and fast algorithm is the key to the success of a bioinformatics tool.

Alexey Gurevich, Head of Natural Products Discovery, the Center for Algorithmic Biotechnology

‘However, this is not its only component. It is of the utmost importance that the created programme is easy to use. It is only like this that it will be in demand by the academic community and will bring real benefits. As in our previous projects to create tools for detecting antibiotics and other biologically active substances, we have made CycloNovo as user-friendly as possible even for scientists with a minimum level of computer literacy. The software application can be launched in just a couple of mouse clicks, and the results of the work can be viewed directly in your Internet browser. You won’t have to install anything on your computer,’ said Alexey Gurevich, candidate of physics and mathematics and head of Natural Products Discovery at the Center for Algorithmic Biotechnology.

Together with Pavel Pevzner, Alexey Gurevich and Andrey Prjibelski, there were other scientists who participated in the work on CycloNano: Bahar Behsaz (a UCSD doctoral student); Fernando Vargas and Professor Larry Smarr (both from the UCSD); Professor Hosein Mohimani from Carnegie Mellon University (the USA); as well as Dr Joshua S Milne and his doctoral student Mark Fisher from the University of Western Australia.

The project was supported by grants from: the Russian Science Foundation; UCSD; US National Institutes of Health; the Australian Research Council; and the Government of Australia.