Machine learning to predict clinical remission in depressed patients after acute phase selective serotonin reuptake inhibitor treatment.

OBJECTIVE: Selective serotonin reuptake inhibitors (SSRIs) are suggested as the first-line treatment for patients with major depressive disorder (MDD), but the remission rate is unsatisfactory. We aimed to establish machine learning models and explore variables available at baseline to predict the 8-week outcome among patients taking SSRIs. METHODS: Data from 400 patients were used to build machine learnings. The last observation carried forward approach was used to determine the remitter/non-remitter status of the patients at week 8. Using least absolute shrinkage and selection operator (LASSO) to select features, we built 4 different machine learning algorithms including gradient boosting decision tree, support vector machine (SVM), random forests, and logistic regression with five-fold cross-validation. Then, we adopted Shapley additive explanations (SHAP) values to interpret the model output. RESULTS: The remission rate is 67.8%. We obtained 78 features from the baseline characteristics, including 25 sociodemographic characteristics, 31 clinical features, 15 psychological traits and 7 neurocognitive functions, and 13 of these features were selected to establish SVM. The accuracy of the SVM prediction is 74.49%, reaching an average area under the curve of 0.734±0.043. The sensitivity is 0.899±0.038 with a positive predictive value of 0.776±0.028. The specificity is 0.422±0.091 with a negative predictive value of 0.674±0.086. According to the SHAP values, neurocognitive functions and anxiety and hypochondriasis symptoms were important predictors. CONCLUSION: Our study supports the utilization of machine learning approaches with inexpensive and highly accessible variables to accurately predict the 8-week treatment outcome of SSRIs in patients with MDD.

Characterization of novel neutralizing mouse monoclonal antibody JM1-24-3 developed against MUC18 in metastatic melanoma.

BACKGROUND: MUC18 is a glycoprotein highly expressed on the surface of melanoma and other cancers which promotes tumor progression and metastasis. However, its mechanism of action and suitability as a therapeutic target are unknown. METHODS: A monoclonal antibody (mAb) (JM1-24-3) was generated from metastatic melanoma tumor live cell immunization, and high-throughput screening identified MUC18 as the target. RESULTS: Analysis of molecular interactions between MUC18 and JM1-24-3 revealed that the downstream signaling events depended on binding of the mAb to a conformational epitope on the extracellular domain of MUC18. JM1-24-3 inhibited melanoma cell proliferation, migration and invasion in vitro and reduced tumor growth and metastasis in vivo. CONCLUSION: These results confirm that MUC18 is mechanistically important in melanoma growth and metastasis, suggest that the MUC18 epitope identified is a promising therapeutic target, and that the JM1-24-3 mAb may serve as the basis for a potential therapeutic agent.

Discovery of the improved antagonistic prolactin variants by library screening.

Prolactin (PRL), a potent growth stimulator of the mammary epithelium, has been suggested to be a factor contributing to the development and progression of breast and prostate cancer. Several PRL receptor (PRLR) antagonists have been identified in the past decades, but their in vivo growth inhibitory potency was restricted by low receptor affinity, rendering them pharmacologically unattractive for clinical treatment. Thus, higher receptor affinity is essential for the development of improved PRLR antagonistic variants with improved in vivo potency. In this study, we generated Site 1 focused protein libraries of human G129R-PRL mutants and screened for those with increased affinity to the human PRLR. By combining the mutations with enhanced affinities for PRLR, we identified a novel G129R-PRL variant with mutations at Site 1 that render nearly 50-fold increase in the antagonistic potency in vitro.

Topological analysis of integral membrane constituents of prokaryotic ABC efflux systems.

The ATP binding cassette (ABC) superfamily consists of dozens of families of transport systems, each of which catalyzes uptake or efflux of a specific type of molecule using ATP hydrolysis to energize transport. While all of the ATP hydrolyzing subunits in the superfamily are homologous, a monophyletic origin of the integral membrane constituents is not established. We have identified a subset of these transmembrane proteins that have a basal unit of four transmembrane alpha-helical segments (TMSs) with a large extracytoplasmic domain between TMSs 1 and 2. These homologues were found to exhibit 4, 8 or 10 putative TMSs per polypeptide chain. The two larger topological types exhibit a 4 TMS repeat element resulting from an internal gene duplication event, and the 10 TMS proteins display an extra two putative TMSs between the two repeat units. Rare intragenic deletions in these homologues gave rise to truncated proteins lacking the extracytoplasmic domain, and some phylogenetic clusters of the 4 TMS membrane proteins (but not the 8 or 10 TMS proteins) are fused N-terminal (never C-terminal) to ATP hydrolyzing domains. Bioinformatic analyses lead to the suggestion that in the larger homologues, the second repeat units are more important for function than the first repeat units. Operon analyses suggest that the 4 TMS proteins form heterodimeric complexes while the 8 and 10 TMS proteins incorporate the equivalent of these complexes into single integral membrane polypeptide chains. Different gene compositions of the operons encoding the 4 versus 8 and 10 TMS homologues suggest that these two structural types of transporters act on different types of substrates and serve dissimilar functions. Significant sequence similarity between the integral membrane constituents of the ABC efflux pumps analyzed here and those of other ABC transporters could not be detected. These studies define the evolutionary pathway taken for the appearance of a subset of ABC transmembrane transport proteins and provide clues regarding their mechanistic and functional characteristics.

Genetic analysis of the HAMP domain of the Aer aerotaxis sensor localizes flavin adenine dinucleotide-binding determinants to the AS-2 helix.

HAMP domains are signal transduction domains typically located between the membrane anchor and cytoplasmic signaling domain of the proteins in which they occur. The prototypical structure consists of two helical amphipathic sequences (AS-1 and AS-2) connected by a region of undetermined structure. The Escherichia coli aerotaxis receptor, Aer, has a HAMP domain and a PAS domain with a flavin adenine dinucleotide (FAD) cofactor that senses the intracellular energy level. Previous studies reported mutations in the HAMP domain that abolished FAD binding to the PAS domain. In this study, using random and site-directed mutagenesis, we identified the distal helix, AS-2, as the component of the HAMP domain that stabilizes FAD binding. AS-2 in Aer is not amphipathic and is predicted to be buried. Mutations in the sequence coding for the contiguous proximal signaling domain altered signaling by Aer but did not affect FAD binding. The V264M residue replacement in this region resulted in an inverted response in which E. coli cells expressing the mutant Aer protein were repelled by oxygen. Bioinformatics analysis of aligned HAMP domains indicated that the proximal signaling domain is conserved in other HAMP domains that are not involved in chemotaxis or aerotaxis. Only one null mutation was found in the coding sequence for the HAMP AS-1 and connector regions, suggesting that these are not active signal transduction sites. We consider a model in which the signal from FAD is transmitted across a PAS-HAMP interface to AS-2 or the proximal signaling domain.

Interactions between the PAS and HAMP domains of the Escherichia coli aerotaxis receptor Aer.

The Escherichia coli energy-sensing Aer protein initiates aerotaxis towards environments supporting optimal cellular energy. The Aer sensor is an N-terminal, FAD-binding, PAS domain. The PAS domain is linked by an F1 region to a membrane anchor, and in the C-terminal half of Aer, a HAMP domain links the membrane anchor to the signaling domain. The F1 region, membrane anchor, and HAMP domain are required for FAD binding. Presumably, alterations in the redox potential of FAD induce conformational changes in the PAS domain that are transmitted to the HAMP and C-terminal signaling domains. In this study we used random mutagenesis and intragenic pseudoreversion analysis to examine functional interactions between the HAMP domain and the N-terminal half of Aer. Missense mutations in the HAMP domain clustered in the AS-2 alpha-helix and abolished FAD binding to Aer, as previously reported. Three amino acid replacements in the Aer-PAS domain, S28G, A65V, and A99V, restored FAD binding and aerotaxis to the HAMP mutants. These suppressors are predicted to surround a cleft in the PAS domain that may bind FAD. On the other hand, suppression of an Aer-C253R HAMP mutant was specific to an N34D substitution with a predicted location on the PAS surface, suggesting that residues C253 and N34 interact or are in close proximity. No suppressor mutations were identified in the F1 region or membrane anchor. We propose that functional interactions between the PAS domain and the HAMP AS-2 helix are required for FAD binding and aerotactic signaling by Aer.

The Aer protein of Escherichia coli forms a homodimer independent of the signaling domain and flavin adenine dinucleotide binding.

In vivo cross-linking between native cysteines in the Aer receptor of Escherichia coli showed dimer formation at the membrane anchor and in the putative HAMP domain. Dimers also formed in mutants that did not bind flavin adenine dinucleotide and in truncated peptides without a signaling domain and part of the HAMP domain.

PAS domain of the Aer redox sensor requires C-terminal residues for native-fold formation and flavin adenine dinucleotide binding.

The Aer protein in Escherichia coli is a membrane-bound, FAD-containing aerotaxis and energy sensor that putatively monitors the redox state of the electron transport system. Binding of FAD to Aer requires the N-terminal PAS domain and residues in the F1 region and C-terminal HAMP domain. The PAS domains of other PAS proteins are soluble in water. To investigate properties of the PAS domain, we subcloned segments of the aer gene from E. coli that encode the PAS domain with and without His6 tags and expressed the PAS peptides in E. coli. The 20-kDa His6-Aer2-166 PAS-F1 fragment was purified as an 800-kDa complex by gel filtration chromatography, and the associating protein was identified by N-terminal sequencing as the chaperone protein GroEL. None of the N-terminal fragments of Aer found in the soluble fraction was released from GroEL, suggesting that these peptides do not fold correctly in an aqueous environment and require a motif external to the PAS domain for proper folding. Consistent with this model, peptide fragments that included the membrane binding region and part (Aer2-231) or all (Aer2-285) of the HAMP domain inserted into the membrane, indicating that they were released by GroEL. Aer2-285, but not Aer2-231, bound FAD, confirming the requirement for the HAMP domain in stabilizing FAD binding. The results raise an interesting possibility that residues outside the PAS domain that are required for FAD binding are essential for formation of the PAS native fold.

Protein secretion systems of Pseudomonas aeruginosa and P fluorescens.

Gram-negative bacteria have evolved numerous systems for the export of proteins across their dual-membrane envelopes. Three of these systems (types I, III and IV) secrete proteins across both membranes in a single energy-coupled step. Four systems (Sec, Tat, MscL and Holins) secrete only across the inner membrane, and four systems [the main terminal branch (MTB), fimbrial usher porin (FUP), autotransporter (AT) and two-partner secretion families (TPS)] secrete only across the outer membrane. We have examined the genome sequences of Pseudomonas aeruginosa PAO1 and Pseudomonas fluorescens Pf0-1 for these systems. All systems except type IV were found in P. aeruginosa, and all except types III and IV were found in P. fluorescens. The numbers of each such system were variable depending on the system and species examined. Biochemical and physiological functions were assigned to these systems when possible, and the structural constituents were analyzed. Available information regarding the mechanisms of transport and energy coupling as well as physiological functions is summarized. This report serves to identify and characterize protein secretion systems in two divergent pseudomonads, one an opportunistic human pathogen, the other a plant symbiont.