Search

Analiza kemijskih prostorov spojin v velikih podatkovnih bazah se že nekaj časa uporablja kot metoda za odkrivanje novih aktivnih molekul. Farmacevtska podjetja nenehno preiskujejo svoje knjižnice spojin v upanju, da bodo odkrila novo spojino vodnico in jo razvila v naslednje uspešno zdravilo. Naš cilj je bil natančno in v širokem obsegu prikazati podobnosti oz. razlike med porazdelitvijo molekularnih lastnosti v kemijskem prostoru spojin, aktivnih na različne mikroorganizme. V ta namen smo analizirali in primerjali kemijski prostor molekul z visoko učinkovitostjo (MIC ⡤ 1 μg/mL), ki so bile testirane na mikroorganizmih ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa in Enterobacter spp.) in E. coli. Vse spojine, skupaj 8.513 edinstvenih struktur, smo našli v podatkovni zbirki ChEMBL, ki je brezplačen in odprt vir informacij. Najprej smo določili Tanimotov koeficient podobnosti med vsemi spojinami v našem vzorcu in vsemi registriranimi zdravili za sistemsko zdravljenje bakterijskih infekcij (ATC razred J01). Kjer je bila vrednost koeficienta podobnosti med zdravili in molekulami našega vzorca najvišja, smo to molekulo uvrstili v antibiotični razred tega zdravila (pod pogojem, da je njuna podobnost presegala nastavljeni prag). Odkrili smo, da večina spojin (~60 %) ni presegala meje 0,5 Tanimotovega koeficienta, kar kaže na potencial za odkrivanje novih in drugačnih protimikrobnih učinkovin, s katerimi bi lahko zdravili okužbe s patogenimi bakterijami. Nadalje smo izvedli statistične teste med vsemi pari mikroorganizmov (t-test za zvezne in chi2 test za diskretne deskriptorje), kjer smo pokazali stopnjo podobnosti znotraj Gram-pozitivnih in znotraj Gram-negativnih bakterij ter raven razlike med obema paroma mikroorganizmov. Velik del raziskave smo posvetili primerjanju oblik molekul pri teh dveh populacijah. Določili smo 3D-obliko vseh molekul v našem vzorcu in s to informacijo izračunali glavne vztrajnostne momente vsake spojine. Izrisali smo trikotne diagrame PMI, kjer smo pokazali, da so spojine, aktivne na Gram-pozitivne mikroorganizme, po obliki bistveno bolj linearne kot pri Gram-negativnih. Delovni tok, s katerim smo primerjali spojine, aktivne na različne mikroorganizme, smo uporabili tudi za primerjanje razlik v kemijskem prostoru med registriranimi protimikrobnimi in drugimi učinkovinami. Našli smo občutne razlike v porazdelitvi molekularnih lastnosti med tema dvema populacijama zdravilnih učinkovin. Ugotovili smo, da je treba odkrivanje in razvoj novih protimikrobnih učinkovin obravnavati na drugačen način kot pri običajnih učinkovinah. The analysis of chemical spaces within large databases has been used for quite some time as a method for searching and finding new active molecules. Corporate screening collections are used constantly to profile activities of their would-be lead compounds to hopefully find the next hit drug. Our goal was to precisely show the difference, or lack thereof, in the distribution of molecular properties in the chemical space of active molecules between populations of microorganisms on a large scale. To this end, we analysed and compared the chemical spaces of highly active compounds (MIC ⡤ 1 μg/mL) tested on ESKAPE microorganisms and E. coli inside the ChEMBL database, a free and open source of information from which a total of 8,513 unique compounds were collected and their properties determined. Firstly, we calculated the Tanimoto coefficients between all molecules from our dataset with all registered antibacterials for systemic use (ATC class J01). We classified all compounds exceeding our threshold of similarity to the antibiotic class of the corresponding best match. We found that most (~60%) molecules did not surpass the set threshold of 0.5, showing the potential of new different types of compounds to be potent agents in combating harmful bacteria. We also conducted statistical tests between all pairs of microorganisms (t-tests for continuous and chi-square tests for nominal properties) showing the degree to which Gram-positive and Gram-negative microorganisms are similar within their own set and dissimilar among each other. A large part of our analysis was dedicated to comparing the shapes of molecules between our populations. We determined the 3D shapes of all molecules in our dataset and with that calculated the principal moments of inertia (PMI) of each compound. By plotting each molecule in the PMI triangle diagram and calculating the difference in the distribution density, we have shown that Gram-positive compounds tend to be significantly more linear in their shape compared to the Gram-negative ones. Finally, we used the tools with which we compared different microorganisms to each other to determine the difference between the chemical space of registered antibiotics and other drugs. Our analysis shows significant deviation of molecular properties between these two populations, demonstrating the need to treat the search and discovery of new antibiotic agents differently to those of typical drugs.