CReM: chemically reasonable mutations

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989592%3A15110%2F20%3A73606929" target="_blank" >RIV/61989592:15110/20:73606929 - isvavai.cz</a>

Výsledek na webu

<a href="https://github.com/DrrDom/crem" target="_blank" >https://github.com/DrrDom/crem</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

CReM: chemically reasonable mutations

Popis výsledku v původním jazyce

CReM is an open-source Python framework to generate chemical structures using a fragment-based approach. It uses historical data about previously synthesized compounds (e.g. ChEMBL) to generate new ones. Known molecules are exhaustively fragmented by breaking acyclic single bonds and the created fragment database is used to generate new structures by replacement of one fragment with others. The major feature is to take into account local context (chemical environment) of fragments and in the course of generation we exchange only fragments occurring in the same chemical context that leads to synthetically more feasible replacements. The local context is encoded by all atoms on the specified maximum distance (radius) from attachment points of a fragment. We demonstrated on a series of ligand-based Guacamol benchmarks that this approach is highly competitive to modern generative models based on neural networks and other state-of-the-art approaches in terms of the quality of generated compounds and their synthetic feasibility. On these benchmarks we also demonstrated that synthetic accessibility of generated compounds can be improved by considering the context of a larger radius and by selection of synthetically more feasible compounds to create fragment databases.

Název v anglickém jazyce

CReM: chemically reasonable mutations

Popis výsledku anglicky

CReM is an open-source Python framework to generate chemical structures using a fragment-based approach. It uses historical data about previously synthesized compounds (e.g. ChEMBL) to generate new ones. Known molecules are exhaustively fragmented by breaking acyclic single bonds and the created fragment database is used to generate new structures by replacement of one fragment with others. The major feature is to take into account local context (chemical environment) of fragments and in the course of generation we exchange only fragments occurring in the same chemical context that leads to synthetically more feasible replacements. The local context is encoded by all atoms on the specified maximum distance (radius) from attachment points of a fragment. We demonstrated on a series of ligand-based Guacamol benchmarks that this approach is highly competitive to modern generative models based on neural networks and other state-of-the-art approaches in terms of the quality of generated compounds and their synthetic feasibility. On these benchmarks we also demonstrated that synthetic accessibility of generated compounds can be improved by considering the context of a larger radius and by selection of synthetically more feasible compounds to create fragment databases.

Druh

R - Software

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/LTARF18013" target="_blank" >LTARF18013: Zvýšení úspěšnosti primárního skríningu biologicky aktivních látek pomocí výpočetních modelů</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2020

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Interní identifikační kód produktu

CReM

Technické parametry

CReM is a cross-platform framework to generate structures of chemical compounds and requires Python 3.6 and RDKit 2019.

Ekonomické parametry

speed up and facilitate de novo design and optimization of compound properties for drug development research; can be integrated in other software as a Python module

IČO vlastníka výsledku

61989592

Název vlastníka

ÚMTM