Digital by design approach to develop a universal deep learning AI architecture for automatic chromatographic peak integration

Chromatographic data processing has garnered attention due to multiple Food and Drug Administration 483 citations and warning letters, highlighting the need for a robust technological solution. The healthcare industry has the potential to greatly benefit from the adoption of digital technologies, but the process of implementing these technologies can be slow and complex. This article presents a “Digital by Design” managerial approach, adapted from pharmaceutical quality by design principles, for designing and implementing an artificial intelligence (AI)-based solution for chromatography peak integration process in the healthcare industry. We report the use of a convolutional neural network model to predict analytical variability for integrating chromatography peaks and propose a potential GxP framework for using AI in the healthcare industry that includes elements on data management, model management, and human-in-the-loop processes. The component on analytical variability prediction has a great potential to enable Industry 4.0 objectives on real-time release testing, automated quality control, and continuous manufacturing.     

Rational design, synthesis and evaluation of (6aR( *),11bS( *))-1-(4-fluorophenyl)-4-{7-[4-(4-fluorophenyl)-4-oxobutyl]1,2,3,4,6,6a,7,11b,12,12a(RS)-decahydropyrazino[2',1':6,1]pyrido[3,4-b]indol-2-yl}-butan-1-one as a potential neuroleptic agent

In our pursuit to prepare a potent antipsychotic compound, a novel 1,2,3,4,6,6a,7,11b,12,12a-decahydropyrazino[2',1':6,1]pyrido[3,4-b]indole derivative was synthesized which incorporates the butyrophenone substructure twice. This molecule has shown D(1), D(2) and 5-HT(2A) receptor blocking activity where the ratio pK(i) (5-HT(2A)) to pK(i) (D(2)) is 1.42 better than risperidone (1.15). It blocks amphetamine induced hyperactivity/stereotypy and secondary conditioned avoidance responses in rodents at lower doses than those required for the neuroleptic drugs haloperidol and centbutindole (biriperone).     

Stromal cell-associated expression of kallikrein-related peptidase 6 (KLK6) indicates poor prognosis of ovarian cancer patients

Several members of the human kallikrein-related peptidase family, including KLK6, are up-regulated in ovarian cancer. High KLK6 mRNA or protein expression, measured by quantitative polymerase chain reaction and enzyme-linked immunoassay, respectively, was previously found to be associated with a shortened overall and progression-free survival (OS and PFS, respectively). In the present study, we aimed at analyzing KLK6 protein expression in ovarian cancer tissue by immunohistochemistry. Using a newly developed monospecific polyclonal antibody, KLK6 immunoexpression was initially evaluated in normal tissues. We observed strong staining in the brain and moderate staining in the kidney, liver, and ovary, whereas the pancreas and the skeletal muscle were unreactive, which is in line with previously published results. Next, both tumor cell- and stromal cell-associated KLK6 immunoexpression were analyzed in tumor tissue specimens of 118 ovarian cancer patients. In multivariate Cox regression analysis, only stromal cell-associated expression, besides the established clinical parameters FIGO stage and residual tumor mass, was found to be statistically significant for OS and PFS [high vs. low KLK6 expression; hazard ratio (HR), 1.92; p=0.017; HR, 1.80; p=0.042, respectively]. These results indicate that KLK6 expressed by stromal cells may considerably contribute to the aggressiveness of ovarian cancer.     

Characterization of the Entamoeba histolytica ornithine decarboxylase-like enzyme

Background

The polyamines putrescine, spermidine, and spermine are organic cations that are required for cell growth and differentiation. Ornithine decarboxylase (ODC), the first and rate-limiting enzyme in the polyamine biosynthetic pathway, is a highly regulated enzyme.

Methodology and Results

To use this enzyme as a potential drug target, the gene encoding putative ornithine decarboxylase (ODC)-like sequence was cloned from Entamoeba histolytica, a protozoan parasite causing amoebiasis. DNA sequence analysis revealed an open reading frame (ORF) of ∼1,242 bp encoding a putative protein of 413 amino acids with a calculated molecular mass of 46 kDa and a predicted isoelectric point of 5.61. The E. histolytica putative ODC-like sequence has 33% sequence identity with human ODC and 36% identity with the Datura stramonium ODC. The ORF is a single-copy gene located on a 1.9-Mb chromosome. The recombinant putative ODC protein (48 kDa) from E. histolytica was heterologously expressed in Escherichia coli. Antiserum against recombinant putative ODC protein detected a band of anticipated size ∼46 kDa in E. histolytica whole-cell lysate. Difluoromethylornithine (DFMO), an enzyme-activated irreversible inhibitor of ODC, had no effect on the recombinant putative ODC from E. histolytica. Comparative modeling of the three-dimensional structure of E. histolytica putative ODC shows that the putative binding site for DFMO is disrupted by the substitution of three amino acids—aspartate-332, aspartate-361, and tyrosine-323—by histidine-296, phenylalanine-305, and asparagine-334, through which this inhibitor interacts with the protein. Amino acid changes in the pocket of the E. histolytica enzyme resulted in low substrate specificity for ornithine. It is possible that the enzyme has evolved a novel substrate specificity.

Conclusion

To our knowledge this is the first report on the molecular characterization of putative ODC-like sequence from E. histolytica. Computer modeling revealed that three of the critical residues required for binding of DFMO to the ODC enzyme are substituted in E. histolytica, resulting in the likely loss of interactions between the enzyme and DFMO.     

MicroRNA regulation of human protease genes essential for influenza virus replication

Influenza A virus causes seasonal epidemics and periodic pandemics threatening the health of millions of people each year. Vaccination is an effective strategy for reducing morbidity and mortality, and in the absence of drug resistance, the efficacy of chemoprophylaxis is comparable to that of vaccines. However, the rapid emergence of drug resistance has emphasized the need for new drug targets. Knowledge of the host cell components required for influenza replication has been an area targeted for disease intervention. In this study, the human protease genes required for influenza virus replication were determined and validated using RNA interference approaches. The genes validated as critical for influenza virus replication were ADAMTS7, CPE, DPP3, MST1, and PRSS12, and pathway analysis showed these genes were in global host cell pathways governing inflammation (NF-κB), cAMP/calcium signaling (CRE/CREB), and apoptosis. Analyses of host microRNAs predicted to govern expression of these genes showed that eight miRNAs regulated gene expression during virus replication. These findings identify unique host genes and microRNAs important for influenza replication providing potential new targets for disease intervention strategies.     

Structural insight of dopamine β-hydroxylase, a drug target for complex traits, and functional significance of exonic single nucleotide polymorphisms

Background

Human dopamine β-hydroxylase (DBH) is an important therapeutic target for complex traits. Several single nucleotide polymorphisms (SNPs) have also been identified in DBH with potential adverse physiological effect. However, difficulty in obtaining diffractable crystals and lack of a suitable template for modeling the protein has ensured that neither crystallographic three-dimensional structure nor computational model for the enzyme is available to aid rational drug design, prediction of functional significance of SNPs or analytical protein engineering.

Principal Findings

Adequate biochemical information regarding human DBH, structural coordinates for peptidylglycine alpha-hydroxylating monooxygenase and computational data from a partial model of rat DBH were used along with logical manual intervention in a novel way to build an in silico model of human DBH. The model provides structural insight into the active site, metal coordination, subunit interface, substrate recognition and inhibitor binding. It reveals that DOMON domain potentially promotes tetramerization, while substrate dopamine and a potential therapeutic inhibitor nepicastat are stabilized in the active site through multiple hydrogen bonding. Functional significance of several exonic SNPs could be described from a structural analysis of the model. The model confirms that SNP resulting in Ala318Ser or Leu317Pro mutation may not influence enzyme activity, while Gly482Arg might actually do so being in the proximity of the active site. Arg549Cys may cause abnormal oligomerization through non-native disulfide bond formation. Other SNPs like Glu181, Glu250, Lys239 and Asp290 could potentially inhibit tetramerization thus affecting function.

Conclusions

The first three-dimensional model of full-length human DBH protein was obtained in a novel manner with a set of experimental data as guideline for consistency of in silico prediction. Preliminary physicochemical tests validated the model. The model confirms, rationalizes and provides structural basis for several biochemical data and claims testable hypotheses regarding function. It provides a reasonable template for drug design as well.     

A Novel Spectral Annotation Strategy Streamlines Reporting of Mono-ADP-ribosylated Peptides Derived from Mouse Liver and Spleen in Response to IFN-γ

Mass-spectrometry-enabled ADP-ribosylation workflows are developing rapidly, providing researchers a variety of ADP-ribosylome enrichment strategies and mass spectrometric acquisition options. Despite the growth spurt in upstream technologies, systematic ADP-ribosyl (ADPr) peptide mass spectral annotation methods are lacking. HCD-dependent ADP-ribosylome studies are common, but the resulting MS2 spectra are complex, owing to a mixture of b/y-ions and the m/p-ion peaks representing one or more dissociation events of the ADPr moiety (m-ion) and peptide (p-ion). In particular, p-ions that dissociate further into one or more fragment ions can dominate HCD spectra but are not recognized by standard spectral annotation workflows. As a result, annotation strategies that are solely reliant upon the b/y-ions result in lower spectral scores that in turn reduce the number of reportable ADPr peptides. To improve the confidence of spectral assignments, we implemented an ADPr peptide annotation and scoring strategy. All MS2 spectra are scored for the ADPr m-ions, but once spectra are assigned as an ADPr peptide, they are further annotated and scored for the p-ions. We implemented this novel workflow to ADPr peptides enriched from the liver and spleen isolated from mice post 4 h exposure to systemic IFN-γ. HCD collision energy experiments were first performed on the Orbitrap Fusion Lumos and the Q Exactive, with notable ADPr peptide dissociation properties verified with CID (Lumos). The m-ion and p-ion series score distributions revealed that ADPr peptide dissociation properties vary markedly between instruments and within instrument collision energy settings, with consequences on ADPr peptide reporting and amino acid localization. Consequentially, we increased the number of reportable ADPr peptides by 25% (liver) and 17% (spleen) by validation and the inclusion of lower confidence ADPr peptide spectra. This systematic annotation strategy will streamline future reporting of ADPr peptides that have been sequenced using any HCD/CID-based method.