Exploiting pretrained biochemical language models for targeted drug design.

MOTIVATION: The development of novel compounds targeting proteins of interest is one of the most important tasks in the pharmaceutical industry. Deep generative models have been applied to targeted molecular design and have shown promising results. Recently, target-specific molecule generation has been viewed as a translation between the protein language and the chemical language. However, such a model is limited by the availability of interacting protein-ligand pairs. On the other hand, large amounts of unlabelled protein sequences and chemical compounds are available and have been used to train language models that learn useful representations. In this study, we propose exploiting pretrained biochemical language models to initialize (i.e. warm start) targeted molecule generation models. We investigate two warm start strategies: (i) a one-stage strategy where the initialized model is trained on targeted molecule generation and (ii) a two-stage strategy containing a pre-finetuning on molecular generation followed by target-specific training. We also compare two decoding strategies to generate compounds: beam search and sampling. RESULTS: The results show that the warm-started models perform better than a baseline model trained from scratch. The two proposed warm-start strategies achieve similar results to each other with respect to widely used metrics from benchmarks. However, docking evaluation of the generated compounds for a number of novel proteins suggests that the one-stage strategy generalizes better than the two-stage strategy. Additionally, we observe that beam search outperforms sampling in both docking evaluation and benchmark metrics for assessing compound quality. AVAILABILITY AND IMPLEMENTATION: The source code is available at https://github.com/boun-tabi/biochemical-lms-for-drug-design and the materials (i.e., data, models, and outputs) are archived in Zenodo at https://doi.org/10.5281/zenodo.6832145. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

5-Fluoro/(trifluoromethoxy)-2-indolinone derivatives with anti-interleukin-1 activity.

The pro-inflammatory cytokine interleukin-1 (IL-1) drives the pathogenesis of several inflammatory diseases. Recent studies have revealed that 2-indolinones can modulate cytokine responses. Therefore, we screened several 2-indolinone derivatives in preliminary studies to develop agents with anti-IL-1 activity. First, the putative efficacies and binding interactions of 2-indolinones were evaluated by docking studies. Second, previously synthesized 5-fluoro/(trifluoromethoxy)-1H-indole-2,3-dione 3-(4-phenylthiosemicarbazones) (compounds 47-69) which had the highest inhibitory effect in the screening were evaluated for inhibitory effects on the IL-1 receptor (IL-1R). Compounds 52 (IC50 = 0.09 µM) and 65 (IC50 = 0.07 µM) were selected as lead compounds for the subsequent synthesis of new derivatives. The novel 5-fluoro/(trifluoromethoxy)-1H-indole-2,3-dione 3-(4-phenylthiosemicarbazones) (compounds 70-116) were designed, synthesized, and in vitro studies were completed. The compounds 76, 78, 81, 91, 100, 105, and 107 tested showed nontoxic inhibitory effects on IL-1R-dependent responses in the range of 0.01-0.06 µM and stronger than the lead compounds 52 and 65. In vitro and in silico findings showed that compounds 78 (IC50 = 0.01 µM) and 81 (IC50 = 0.02 µM) had the strongest IL-1R inhibitory effects and the most favorable drug-like properties. Molecular modeling studies of the compounds 78 and 81 were carried out to determine the possible binding interactions at the active site of the IL-1R.

New 1H-indole-2,3-dione 3-thiosemicarbazones with 3-sulfamoylphenyl moiety as selective carbonic anhydrase inhibitors.

1-Methyl/ethyl/benzyl-5-(un)substituted 1H-indole-2,3-diones (2, 3, and 4) were synthesized by reaction of 5-(un)substituted 1H-indole-2,3-diones (1) with methyl iodide, ethyl chloride, and benzyl bromide. (3-Sulfamoylphenyl)isothiocyanate (6) was obtained by the treatment of 3-aminobenzenesulfonamide (5) with thiophosgene. Compound 6 was reacted with hydrazine to yield 4-(3-sulfamoylphenyl)thiosemicarbazide (7). Novel 1-(un)substituted/methyl/ethyl/benzyl-5-(un)substituted 1H-indole-2,3-dione 3-[4-(3-sulfamoylphenyl)thiosemicarbazone] derivatives (8-11) were prepared by condensation of 7 and 1-4. The structures of the synthesized compounds were confirmed by elemental analysis and spectral data. Inhibition of the widely distributed cytosolic off-targets human carbonic anhydrases (hCAs) I and II, and two tumor-associated membrane-bound isoforms (hCAs IX and XII), by 8-11 was investigated. The hCA II inhibitory effects of all tested compounds were in the subnanomolar to low nanomolar levels (Ki = 0.32-83.3 nM), and generally high selectivity for hCA II isoenzyme over hCA I, IX, and XII isoenzymes was observed. The strongest inhibitors of hCA II, 1-benzyl-5-(trifluoromethoxy)-substituted 11c (Ki = 0.32 nM) and 1-ethyl-5-chloro-substituted 10e (Ki = 0.35 nM), were docked within the enzyme active site. Molecular modeling studies with the most effective hCA IX and XII inhibitors were also carried out.

Synthesis, molecular modeling and antiviral activity of novel 5-fluoro-1H-indole-2,3-dione 3-thiosemicarbazones.

In this work, novel 5-fluoro-1-methyl/ethyl-1H-indole-2,3-dione 3-[4-(substituted phenyl)-thiosemicarbazones] 6a-n and 7a-n were synthesized. The antiviral effects of the compounds were tested against HSV-1 (KOS), HSV-2 (G) HSV-1 TK- KOS ACVr and VV in HEL cell cultures using acyclovir and ganciclovir as standards, and Coxsackie B4 virus in Vero cell cultures using ribavirin and mycophenolic acid as standards. R2 ethyl substituted 7 derivatives were found effective against viruses tested. R1 4-CF3 substituted 7d, R1 4-OCH3 substituted 7 g and R1 3-Cl substituted 7 l showed activity against HSV-1 (KOS), HSV-2 (G) HSV-1 TK- KOS ACVr and VV. Whereas only R1 4-Br substituted 7n has selective activity against coxsackie B4 virus. Molecular modelingstudies of 7d and 7l were performed to determine binding side on HSV-1 glycoprotein B and D, HSV-2 glycoprotein B structures.

Novel 2-indolinones containing a sulfonamide moiety as selective inhibitors of candida ß-carbonic anhydrase enzyme.

Inhibition of the ß-carbonic anhydrase (CA, EC 4.2.1.1) from pathogenic Candida glabrata (CgNce103) by 1H-indole-2,3-dione 3-[N-(4-sulfamoylphenyl)thiosemicarbazones] 4a-m was investigated. All the compounds were found to be potent inhibitors of CgNce103, with inhibition constants in the range of 6.4-63.9 nM. The 5,7-dichloro substituted derivative 4l showed the most effective inhibition (KI of 6.4 nM) as well as the highest selectivity for inhibiting CgNce103 over the cytosolic human (h) isoforms hCA I and II. A possible binding interaction of compound 4l within the active site of CgNce103 has been proposed based on docking studies.

Novel sulfonamide-containing 2-indolinones that selectively inhibit tumor-associated alpha carbonic anhydrases.

Human carbonic anhydrases IX and XII are upregulated in many tumors and form a novel target for new generation anticancer drugs. Here we report the synthesis of novel 2-indolinone derivatives with the sulfonamide group as a zinc binding moiety. Enzyme inhibition assays confirmed that the compounds showed selectivity against hCA IX and XII over the widely distributed off-targets hCA I and II. Molecular modelling studies were performed to suggest modes of binding for these compounds.

Isatin analogs as novel inhibitors of Candida spp. ß-carbonic anhydrase enzymes.

Enzyme inhibition data of structurally novel isatin-containing sulfonamides were determined for two carbonic anhydrases (CAs, EC 4.2.1.1) from pathogenic Candida species (CaNce103 from C. albicans and CgNce103 from C. glabrata). The compounds show KI values in the low nanomolar range for the fungal CAs, while they have significantly higher KI values for the human CAs. Homology models were constructed for the CaNce103 and CgNce103 and subsequently the ligands were docked into these models to rationalize their enzyme inhibitory properties.

Discovery of novel isatin-based sulfonamides with potent and selective inhibition of the tumor-associated carbonic anhydrase isoforms IX and XII.

A series of 2/3/4-[(2-oxo-1,2-dihydro-3H-indol-3-ylidene)amino]benzenesulfonamides, obtained from substituted isatins and 2-, 3- or 4-aminobenzenesulfonamide, showed low nanomolar inhibitory activity against the tumor associated carbonic anhydrase (CA, EC 4.2.1.1) isoforms IX and XII - recently validated antitumor drug targets, being much less effective as inhibitors of the off-target cytosolic isoforms CA I and II.

New spiroindolinones bearing 5-chlorobenzothiazole moiety.

In this study, 5-chloro-3H-spiro-[1,3-benzothiazole-2,3'-indole]-2'(1'H)-one derivatives 3a-l were synthesized by the reaction of 1H-indole-2,3-diones 1a-l with 2-amino-4-chlorothiophenol 2 in ethanol. 3a-l were tested for their abilities to inhibit lipid peroxidation (LP), scavenge DPPH(•) and ABTS(•+) radicals, and to reduce Fe(3+) to Fe(2+). Most of the tested compounds exhibited potent scavenging activities against ABTS(•+) radical, reducing powers and strong inhibitory capacity on LP. 3 a, 3 d, 3 e, 3h, 3 j and 3 k chosen as prototypes were evaluated in the National Cancer Institute's in vitro primary anticancer assay. The greatest growth inhibitions were observed against a non-small cell lung cancer cell line HOP-92 for R1-fluoro substituted 3 d and a renal cancer cell line RXF-393 for R-chloro substituted 3 e in the primary screen.

Novel 2-indolinone derivatives as promising agents against respiratory syncytial and yellow fever viruses.

Background: A vaccine or antiviral drug for respiratory syncytial virus (RSV) infections and a specific antiviral drug for yellow fever virus (YFV) infections has not yet been developed. Method: In this study, 2-indolinone-based N-(4-sulfamoylphenyl)hydrazinecarbothioamides were synthesized. Along with these new compounds, previously synthesized 2-indolinone-based N-(3-sulfamoylphenyl)hydrazinecarbothioamides were evaluated against various DNA and RNA viruses. Results: Some 2-indolinone compounds exhibited nontoxic and selective antiviral activities against RSV and YFV. Halogen substitution at the indole ring increased the anti-RSV activities. Moreover, 1-benzyl and 5-halogen or nitro-substituted compounds were the most effective compounds against YFV. Conclusion: Generally, the 3-sulfonamide-substituted compounds were determined to be more effective than 4-sulfonamide-substituted compounds against RSV and YFV.

Synthesis and molecular docking studies of 5-trifluoromethoxy-2-indolinones as cholinesterase dual inhibitors.

Background: In Alzheimer's disease, butyrylcholinesterase (BuChE) activity gradually increases, while acetylcholinesterase (AChE) activity decreases or remains unchanged. Dual inhibitors have important roles in regulation of synaptic acetylcholine levels and progression of Alzheimer's disease. Methods: 1-(Thiomorpholin-4-ylmethyl)/benzyl-5-trifluoromethoxy-2-indolinones (6-7) were synthesized. AChE and BuChE inhibitory effects were investigated with Ellman's method. Molecular docking studies were performed for analyzing the possible binding interactions at active sites. Results: Compound 6g was the strongest inhibitor against both AChE (Ki = 0.35 µM) and BuChE (Ki = 0.53 µM). It showed higher inhibitory effects than both donepezil and galantamine. Moreover, compound 7m had a higher inhibitory effect than galantamine and the effect was comparable to that of donepezil against both AChE (Ki = 0.69 µM) and BuChE (Ki = 0.95 µM). Conclusion: The benzyl substitution compared with 1-(thiomorpholin-4-ylmethyl) group significantly increased both AChE and BuChE inhibitory effects.

1H-Indole-2,3-dione 3-thiosemicarbazones carrying a 4-sulfamoylphenyl moiety with selective antiviral activity against reovirus-1.

1H-indole-2,3-dione 3-[4-(4-sulfamoylphenyl)thiosemicarbazones] (6a-j) were evaluated against Para-influenza-3, Reovirus-1, Sindbis, Coxsackie B4 and Punto Toro viruses. New 1-methyl-1H-indole-2,3-dione 3-[4-(4-sulfamoylphenyl)thiosemicarbazones] (7a-c) were synthesized to evaluate the contribution of methyl substitution at position 1- of the indole ring to antiviral activity. The test results showed that compounds 5-trifluoromethoxy- substituted 6c (EC50: 2-9 µM) and 5-bromo- substituted 6f (EC50: 2-3 µM) have non-toxic selective antiviral activity while not all standards are active against Reovirus-1. Molecular docking studies of 6c and 6f were carried out to determine the possible binding positions with Reovirus-1. Trifluoromethoxy and bromine substitutions at position 5- of the indole ring provided selective antiviral activity, while methyl substitution at position 1- of the indole ring significantly decreased the activity and increased toxicity against Reovirus-1.

Exploring data-driven chemical SMILES tokenization approaches to identify key protein-ligand binding moieties.

Machine learning models have found numerous successful applications in computational drug discovery. A large body of these models represents molecules as sequences since molecular sequences are easily available, simple, and informative. The sequence-based models often segment molecular sequences into pieces called chemical words, analogous to the words that make up sentences in human languages, and then apply advanced natural language processing techniques for tasks such as de novo drug design, property prediction, and binding affinity prediction. However, the chemical characteristics and significance of these building blocks, chemical words, remain unexplored. To address this gap, we employ data-driven SMILES tokenization techniques such as Byte Pair Encoding, WordPiece, and Unigram to identify chemical words and compare the resulting vocabularies. To understand the chemical significance of these words, we build a language-inspired pipeline that treats high affinity ligands of protein targets as documents and selects key chemical words making up those ligands based on tf-idf weighting. The experiments on multiple protein-ligand affinity datasets show that despite differences in words, lengths, and validity among the vocabularies generated by different subword tokenization algorithms, the identified key chemical words exhibit similarity. Further, we conduct case studies on a number of target to analyze the impact of key chemical words on binding. We find that these key chemical words are specific to protein targets and correspond to known pharmacophores and functional groups. Our approach elucidates chemical properties of the words identified by machine learning models and can be used in drug discovery studies to determine significant chemical moieties.

Bioisosterism: 1,2,4-Oxadiazole Rings.

Although studies in drug discovery have gained momentum in recent years, the conversion of drugs in use today into less toxic derivatives with pharmacologically superior properties is still of great importance in drug research. Bioisosterism facilitates the conversion of drugs into derivatives that present more positive pharmacological and toxicological profiles by changing existing groups in the drug structure within the framework of certain criteria that have been expanded today. The 1,2,4-oxadiazole ring is used as a bioisostere for ester and amide groups due to its resistance to hydrolysis. However, this ring is not limited to esters and amides, but can also be used as a bioisostere for other functional groups. In this review, cases in which the 1,2,4-oxadiazole ring is used as a bioisostere for various functional groups are discussed. Herein we shed light on 1,2,4-oxadiazole bioisosterism in the development of new drug candidates and in enhancing the pharmacological profiles of currently available drugs.

Synthesis and molecular modeling studies of 1-benzyl-2-indolinones as selective AChE inhibitors.

Background: Possible bioisosteres can be developed by replacing the 1-indanone ring (one of three pharmacophore groups) of donepezil with an indoline ring. As H2S donors, thioamide, thiocarbamate and thiourea groups are also critically important. Materials & methods: The 1-benzyl-2-indolinones 6a-n were designed using molecular modeling and synthesized, and their acetylcholinesterase and butyrylcholinesterase inhibitory effects were then investigated. Results: The compounds 6h (inhibition constant [Ki] = 0.22 µM; selectivity index [SI] = 26.22), 6i (Ki = 0.24 µM; SI = 25.83), 6k (Ki = 0.22 µM; SI = 28.31) and 6n (Ki = 0.21 µM; SI = 27.14) were approximately twofold more effective against and >12-fold more selective for acetylcholinesterase compared with donepezil (Ki = 0.41 µM; SI = 2.12). Analysis of molecular dynamics simulations with compounds 6k and 6n indicated that the preferred binding might be at allosteric binding pocket 4 of the enzyme. Conclusion: Benzyl substitution at the 1-position of the indole ring significantly increased potency and selectivity.

Selective Cytotoxic Effects of 5-Trifluoromethoxy-1H-indole-2,3-dione 3-Thiosemicarbazone Derivatives on Lymphoid-originated Cells.

AIM: The present study aims to identify the anticancer effect of novel 1H-indole-2,3-dione 3- thiosemicarbazone derivatives. These compounds could be promising anticancer agents in leukemia treatment. BACKGROUND: Conventional chemotherapeutic agents accumulate in both normal and tumor cells due to nonspecificity. For effective cancer treatment, new drugs need to be developed to make chemotherapeutics selective for cancer cells. The ultimate goal of cancer treatment is to reduce systemic toxicity and improve the quality of life. METHODS: In this study, the anticancer effects of 5-trifluoromethoxy-1H-indole-2,3-dione 3-thiosemicarbazone derivatives (A-L) were investigated in chronic myelogenous leukemia K562, Burkitt's lymphoma P3HR1, acute promyelocytic leukemia HL60 cells, and vincristine-resistant sublines of K562 and P3HR1 cells. Additionally, the compounds were tested on lymphoid-derived cells from ALL patients. In order to investigate the particular mechanism of death caused by the cytotoxic effects of the compounds, immunohistochemical caspase 3 staining was performed in P3HR1 cells, and the resulting apoptotic activities were demonstrated. RESULTS: All tested compounds have been found to have cytotoxic effects against lymphoma cells at submicromolar concentrations (IC50= 0.89-1.80 µM). Most compounds show significant selectivity for the P3HR1 and P3HR1 Vin resistance. The most effective and selective compound is 4-bromophenyl substituted compound I (IC50=0.96 and 0.89 µM). Cyclohexyl and benzyl substituted compounds D and E have also been found to have cytotoxic effects against K562 cell lines (IC50=2.38 µM), while the allyl substituted compound C is effective on all cell lines (IC50=1.13-2.21 µM). 4-Fluorophenyl substituted F compound has been observed to be effective on all cells (IC50=1.00-2.41 µM) except K562 cell. Compound C is the only compound that shows inhibition of HL-60 cells (IC50= 1.13 µM). Additionally, all compounds exhibited cytotoxic effects on lymphoidderived cells at 1µM concentration. These results are in accordance with the results obtained in lymphoma cells. CONCLUSION: All compounds tested have submicromolar concentrations of cytotoxic effects on cells. These compounds hold potential for use in future treatments of leukemia.

Synthesis, molecular modeling and cholinesterase inhibitory effects of 2-indolinone-based hydrazinecarbothioamides.

Background: 2-Indolinone-based hydrazinecarbothioamides carrying a 3-phenylsulfonamide moiety (7-9) were designed by replacement of donepezil's pharmacophore group indanone with a 2-indolinone ring. Method: Compounds 7-9 were synthesized by reaction of N-(3-sulfamoylphenyl)hydrazinecarbothioamide (6) with 1H-indolin-2,3-diones (1-3). Acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE) inhibitory effects of compounds 7-9 were assayed. Molecular modeling studies of 5-chloro-1,7-dimethyl-substituted compound 8e were carried out to determine the possible binding interactions at the active site of AChE. Results: Compound 8e showed the strongest inhibition against AChE (Ki = 0.52 ± 0.11 µM) as well as the highest selectivity (SI = 37.69). The selectivity for AChE over BuChE of compound 8e was approximately 17-times higher than donepezil and 26-times higher than galantamine. Conclusion: Further development of compounds 7-9 may present new promising agents for Alzheimer's treatment.

5-Chloro-benzothia-zole-2-spiro-3'-indolin-2'-one.

The title compound, C(14)H(9)ClN(2)OS, crystallizes with two unique mol-ecules, A and B, in the asymmetric unit. The five-membered rings of the benzothia-zole groups in both mol-ecules adopt an envelope conformation [puckering parameters: q(2) = 0.242 (1) Šand ϕ(2) = 217.5 (4)° for A, and q(2) = 0.234 (1) Šand ϕ(2) = 37.7 (4)° for B]. The five-membered rings of the indolinone groups in both mol-ecules are also not planar, with a twisted conformation [puckering parameters are q(2) = 0.112 (2) Šand ϕ(2) = 126.3 (8)° for A, and q(2) = 0.108 (2) Šand ϕ(2) = 306.4 (9)° for B]. In the crystal structure, there are inter-molecular N-H⋯O, N-H⋯S and C-H⋯O hydrogen-bonding inter-actions, forming the layers propagating normal to c.

Synthesis and antituberculosis activity of 5-methyl/trifluoromethoxy-1H-indole-2,3-dione 3-thiosemicarbazone derivatives.

New series of 5-methyl/trifluoromethoxy-1H-indole-2,3-dione 3-thiosemicarbazones 3a-t, 1-methyl-5-methyl/trifluoromethoxy-1H-indole-2,3-dione 3-thiosemicarbazones 4a-y and 5-trifluoromethoxy-1-morpholinomethyl-1H-indole-2,3-dione 3-thiosemicarbazones 5a-m were synthesized. The structures of the synthesized compounds were confirmed by spectral data and elemental analysis. The new 5-methyl/trifluoromethoxy-1H-indole-2,3-dione derivatives, along with previously synthesized 5-methyl-1H-indole-2,3-dione 3-thiosemicarbazones 6a-l, were evaluated for in vitro antituberculosis activity against Mycobacterium tuberculosis H37Rv. 5-Methyl-1H-indole-2,3-dione 3-thiosemicarbazones (3b, 3d, 3f, 6c, 6d, and 6f), 5-trifluoromethoxy-1H-indole-2,3-dione 3-thiosemicarbazones (3q-s) and 5-trifluoromethoxy-1-morpholinomethyl-1H-indole-2,3-dione 3-thiosemicarbazones (5e and 5j-l) were found to be the most potent inhibitors of M. tuberculosis growth described in this study.

Synthesis and primary cytotoxicity evaluation of new 5-nitroindole-2,3-dione derivatives.

A new series of 5-nitro-1H-indole-2,3-dione-3-thiosemicarbazones (3a-k) obtained by condensation of 5-nitro-1H-indole-2,3-dione (1) with N-substituted-thiosemicarbazides (2a-k) were treated with morpholine or piperidine and formaldehyde to yield 1-morpholino/piperidinomethyl-5-nitroindole-2,3-dione-3-thiosemicarbazones (4a-m). The structures of all the compounds were determined by analytical and spectral (IR, 1H-NMR, EIMS) methods. Compounds 3b, 3c, 3f, 3k, 4a, 4c, 4f and 4l chosen as prototypes were evaluated in the National Cancer Institute's 3-cell line, one dose in vitro primary cytotoxicity assay. All the compounds that passed the criteria for activity in this assay were scheduled automatically for evaluation against the full panel of 60 human tumour cell lines at a minimum of five concentrations at 10-fold dilutions. Sulphorhodamine B (SRB) protein assay was used to estimate cell stability or growth. The most active compound was found to be 1-morpholinomethyl-5-nitroindole-2,3-dione-3-N-(chlorophenyl)thiosemicarbazone (4l). This compound demonstrated the most marked effects in the National Cancer Institute's 60 human tumour cell line in vitro screen on a non-small cell lung cancer cell line (HOP-62, log(10)GI(50) value <-8.00) and on leukaemia cell lines (HL-60(TB), log(10)GI(50) value -6.30; MOLT-4, log(10)GI(50) value -6.18).