Are 2D fingerprints still valuable for drug discovery?

Abstract

Recently, molecular fingerprints extracted from three-dimensional (3D) structures using advanced mathematics, such as algebraic topology, differential geometry, and graph theory have been paired with efficient machine learning, especially deep learning algorithms to outperform other methods in drug discovery applications and competitions. This raises the question of whether classical 2D fingerprints are still valuable in computer-aided drug discovery. This work considers 23 datasets associated with four typical problems, namely protein-ligand binding, toxicity, solubility and partition coefficient to assess the performance of eight 2D fingerprints. Advanced machine learning algorithms including random forest, gradient boosted decision tree, single-task deep neural network and multitask deep neural network are employed to construct efficient 2D-fingerprint based models. Additionally, appropriate consensus models are built to further enhance the performance of 2D-fingerprintbased methods. It is demonstrated that 2D-fingerprint-based models perform as well as the state-of-the-art 3D structure-based models for the predictions of toxicity, solubility, partition coefficient and protein-ligand binding affinity based on only ligand information. However, 3D structure-based models outperform 2D fingerprint-based methods in complex-based protein-ligand binding affinity predictions. * Corresponding to Guo-Wei Wei.Molecular fingerprints are one way of encoding the structural features of a molecule. They play a fundamental role in QSAR/QSPR analysis, virtual screening, similarity-based compound search, target molecule ranking, drug ADMET prediction, and other drug discovery processes. Molecular fingerprints are property profiles of a molecule, usually in the form of vectors with each vector element indicating the existence, the degree or the frequency of one particular structure feature. [12][13][14] Various fingerprints have been developed for molecular feature encoding in the past few decades. [15][16][17] Most fingerprints are 2D fingerprints which can be extracted from molecular connection tables without 3D structure information. However, high dimensional fingerprints have also been developed to utilize 3D molecular structure and other information. 18 There are four main categories of 2D fingerprints, namely substructure key-based fingerprints, topological or path-based fingerprints, circular fingerprints, and pharmacophore fingerprints. Substructure key-based fingerprints are bit strings representing the presence of certain substructures or fragments from a given list of structural keys in the compound. Molecular access system (MACCS) 19 is one of the most popular substructure keybased fingerprint methods. Topological or path-based fingerprints are based on analyzing all the fragments of a molecule following a (usually linear) path up to a certain number of bonds, and then hashing every one of these paths to create one fingerprint. The most prominent ones in this category are FP2, 20 Daylight 21 and electrotopological sta...

An Attempt to Boost Molecular Descriptors with Quantum-Derived Features in Prediction of Maximum Emission Wavelengths of Chromophores

“…Only 2D molecular descriptors were utilized to train machine learning models. There are fields that 3D molecular descriptors perform better then 2D ones [20]. Some other applications of machine learning in predictions of organic compounds emission wavelengths were published [21][22].…”

Diversity oriented Deep Reinforcement Learning for targeted molecule generation

“…The first approach, used as the baseline, employs the Extended Connectivity Fingerprint (ECFP) as molecular representation. These bit vectors are widely used in the prediction of physicochemical properties, biological activity or toxicity of chemical compounds [24]. The model output is a real number, which is the estimated pIC 50 .…”

Pushing the limits of solubility prediction via quality-oriented data selection

“…Most 2D descriptors are calculated with absolute accuracy while the 3D descriptors carry the errors of the methodological approximations they have been calculated with (Raevsky et al, 2019). Admitting that the 3D descriptors provide more detailed information, such as atomic distances and energy data of the compounds, there is yet no clear evidence about their impacts on the solubility predictions (Balakin et al 2006;Gao et al, 2020;Yan et al, 2004;Salahinejad et al, 2013). Although a large number of chemical descriptors are available, it is usually preferred to use a modest number of relevant descriptors to avoid redundancy and overfitting issues during the training of ML models (Wang and Hou 2011).…”