Quantitative Structure-Activity Relationship Machine Learning Models and their Applications for Identifying Viral 3CLpro- and RdRp-Targeting Compounds as Potential Therapeutics for COVID-19 and Related Viral Infections.

In response to the ongoing COVID-19 pandemic, there is a worldwide effort being made to identify potential anti-SARS-CoV-2 therapeutics. Here, we contribute to these efforts by building machine-learning predictive models to identify novel drug candidates for the viral targets 3 chymotrypsin-like protease (3CLpro) and RNA-dependent RNA polymerase (RdRp). Chemist-curated training sets of substances were assembled from CAS data collections and integrated with curated bioassay data. The best-performing classification models were applied to screen a set of FDA-approved drugs and CAS REGISTRY substances that are similar to, or associated with, antiviral agents. Numerous substances with potential activity against 3CLpro or RdRp were found, and some were validated by published bioassay studies and/or by their inclusion in upcoming or ongoing COVID-19 clinical trials. This study further supports that machine learning-based predictive models may be used to assist the drug discovery process for COVID-19 and other diseases.

Potential Therapeutic Agents and Associated Bioassay Data for COVID-19 and Related Human Coronavirus Infections.

The COVID-19 pandemic, caused by the novel coronavirus SARS-CoV-2, has led to several million confirmed cases and hundreds of thousands of deaths worldwide. To support the ongoing research and development of COVID-19 therapeutics, this report provides an overview of protein targets and corresponding potential drug candidates with bioassay and structure-activity relationship data found in the scientific literature and patents for COVID-19 or related virus infections. Highlighted are several sets of small molecules and biologics that act on specific targets, including 3CLpro, PLpro, RdRp, S-protein-ACE2 interaction, helicase/NTPase, TMPRSS2, and furin, which are involved in the viral life cycle or in other aspects of the disease pathophysiology. We hope this report will be valuable to the ongoing drug repurposing efforts and the discovery of new therapeutics with the potential for treating COVID-19.

Impact of Chemist-In-The-Loop Molecular Representations on Machine Learning Outcomes.

The development of molecular descriptors is a central challenge in cheminformatics. Most approaches use algorithms that extract atomic environments or end-to-end machine learning. However, a looming question is that how do these approaches compare with the critical eye of trained chemists. The CAS fingerprint engages expert chemists to curate chemical motifs, which they deem could influence bioactivity. In this paper, we benchmark the CAS fingerprint against commonly used fingerprints using a well-established benchmark set of 88 targets. We show that the CAS fingerprint outperforms most of the commonly used molecular fingerprints. Analysis of the CAS fingerprint reveals that experts tend to select features that are rarely reported in the literature, though not all rare features are selected. Our analysis also shows that the CAS fingerprint provides a different source of information compared to other commonly used fingerprints. These results suggest that anthropomorphic insights do have predictive power and highlight the importance of a chemist-in-the-loop approach in the era of machine learning.

Photoaffinity labeling with 8-azidoadenosine and its derivatives: chemistry of closed and opened adenosine diazaquinodimethanes.

The reactive intermediate produced upon photolysis of 8-azidoadenosine was studied by chemical trapping studies, laser flash photolysis with UV-vis and IR detection, and modern computational chemistry. It is concluded that photolysis of 8-azidoadenosine in aqueous solution releases the corresponding singlet nitrene which rapidly tautomerizes to form a closed adenosine diazaquinodimethane in less than 400 fs. A perbenzoylated derivative of 8-azidoadenosine cannot undergo this tautomerization, and instead, it fragments upon photolysis to form an opened adenosine diazaquinodimethane. The singlet nitrene is too short-lived to be observed and, thus, to relax to the lowest triplet state or to become covalently attached to targeted biological macromolecules. The pivotal closed adenosine diazaquinodimethane, the product of nitrene tautomerization, has a lifetime of ca. 1 min or longer in water and in HEPES buffer at ambient temperature. However, this intermediate reacts rapidly with good nucleophiles such as amines, thiols, and phenolates, and significantly more slowly with weak nucleophiles such as alcohols and water. On the basis of these studies, it is clear that the closed adenosine diazaquinodimethane, and not the singlet or triplet nitrene, is the pivotal reactive intermediate involved in photolabeling and cross-linking studies using the 8-azidoadenosine family of photoaffinity labeling reagents.

A study of 2-azido-3,5-dichlorobiphenyl by nano- and picosecond laser flash photolysis and computational methods.

The dynamics of singlet 3,5-dichloro-2-biphenylnitrene and the products of its rearrangement were monitored by pico- and nanosecond laser flash photolysis and the results were consistent with the predictions of DFT and ab initio molecular orbital calculations.

Solvent reorganization controls the rate of proton transfer from neat alcohol solvents to singlet diphenylcarbene.

Femtosecond transient absorption spectroscopy was used to study singlet diphenylcarbene generated by photodissociation of diphenyldiazomethane with a UV pulse at 266 nm. Absorption by singlet diphenylcarbene was detected and characterized for the first time. Similar band shapes were observed in acetonitrile and in cyclohexane with lambda(max) approximately 370 nm. The singlet absorption decays by intersystem crossing to triplet diphenylcarbene at rates that agree with previous measurements. The singlet absorption band is completely formed 1 ps after the pump pulse. It is preceded by a strong and broad absorption band, which is tentatively assigned to excited-state absorption by a singlet diazo excited state. In neat alcohol solvents the growth and decay of the diphenylmethyl cation was observed. This species is formed by proton transfer from an alcohol molecule to singlet diphenylcarbene. Since a shell of solvent molecules surrounds each nascent carbene, the intrinsic rate of protonation in the absence of diffusion could be measured. In methanol, proton transfer occurs with a time constant of 9.0 ps, making this the fastest known intermolecular proton-transfer reaction to carbon. In O-deuterated methanol proton transfer occurs in 15.0 ps. Slower rates were observed in the longer alcohols. The protonation times correlate reasonably well with solvation times in these alcohols, suggesting that solvent fluctuations are the rate-limiting step. In all alcohols studied, the carbocations decay on a somewhat slower time scale to yield diphenylalkyl ethers. In methanol and ethanol the rate of decay is determined by reaction with neutral solvent nucleophiles. There is evidence in 2-propanol that geminate reaction within the initial ion pair is faster than reaction with solvent. No isotope effect was observed for the reaction of the diphenylmethyl carbocation in methanol. Using comparative actinometry the quantum yield of protonation was measured. In methanol, the quantum yield of carbocations reaches a maximum value of 0.18 approximately 18 ps after the pump pulse. According to our analysis, 30% of the photoexcited diazo precursor molecules are eventually protonated. Somewhat lower protonation efficiencies are observed in the other alcohols. Because the primary quantum yield for formation of singlet diphenylcarbene is still unknown, the importance of reaction channels that might exist in addition to protonation cannot be determined at present. Singlet carbenes are powerful, photogenerated bases that open new possibilities for fundamental studies of proton transfer in solution.