Use of machine learning in prediction of granule particle size distribution and tablet tensile strength in commercial pharmaceutical manufacturing.

In the manufacturing of pharmaceutical Oral Solid Dosage (OSD) forms, Particle Size Distribution (PSD) and Tensile Strength (TS) are common in-process tests that are controlled in order to achieve the quality targets of the end-product. The Quality by Design (QbD) concept elaborates process understanding and sufficient controls. However, for older pharmaceutical products upscaled to commercial phase with Quality by Testing (QbT) approach, the knowhow of the product-specific critical parameters is often limited. In this study, two predictive machine learning (ML) models were used for a commercial tablet product, for which historical data of raw materials, production, in-process controls and condition monitoring were available. With the aforementioned data, the aim was to predict the PSD and the TS that indicate the product quality. The feature importance was used to evaluate the parameter importance for the PSD and the TS. Partial dependence, in turn, was used to estimate the parameter impact on the predicted TS. The study illustrates the capability of the ML models to bring additional value for commercial products through the enhanced product-related knowhow.

Long-term performance and stability of molecular shotgun lipidomic analysis of human plasma samples.

The stability of the lipid concentration levels in shotgun lipidomics analysis was tracked over a period of 3.5 years. Concentration levels in several lipid classes, such as phospholipids, were determined in human plasma lipid extracts. Impact of the following factors on the analysis was investigated: sample amount, internal standard amount, and sample dilution factor. Moreover, the reproducibility of lipid profiles obtained in both polarity modes was evaluated. Total number of samples analyzed was approximately 6800 and 7300 samples in negative and positive ion modes, respectively, out of which 610 and 639 instrument control samples were used in stability calculations. The assessed shotgun lipidomics approach showed to be remarkably robust and reproducible, requiring no batch corrections. Coefficients of variation (CVs) of lipid mean concentration measured with optimized analytical parameters were typically less than 15%. The high reproducibility indicated that no lipid degradation occurred during the monitored time period.

Beyond LDL-C lowering: distinct molecular sphingolipids are good indicators of proprotein convertase subtilisin/kexin type 9 (PCSK9) deficiency.

OBJECTIVES: Inhibition of proprotein convertase subtilisin/kexin type 9 (PCSK9) has been proposed to be a potential new therapeutic target for treatment of hypercholesterolaemia. However, little is known about the effects of PCSK9 inhibition on the lipidome. METHODS: We performed molecular lipidomic analyses of plasma samples obtained from PCSK9-deficient mice, and serum of human carriers of a loss-of-function variant in the PCSK9 gene (R46L). RESULTS: In both mouse and man, PCSK9 deficiency caused a decrease in several cholesteryl esters (CE) and short fatty acid chain containing sphingolipid species such as CE 16:0, glucosyl/galactosylceramide (Glc/GalCer) d18:1/16:0, and lactosylceramide (LacCer) d18:1/16:0. In mice, the changes in lipid concentrations were most prominent when animals were given regular chow diet. In man, a number of molecular lipid species was shown to decrease significantly even when LDL-cholesterol was non-significantly reduced by 10% only. Western diet attenuated the lipid lowering potency of PCSK9 deficiency in mice. CONCLUSIONS: Plasma molecular lipid species may be utilized for characterizing novel compounds inhibiting PCSK9 and as sensitive efficacy markers of the PCSK9 inhibition.

A novel method for assigning functional linkages to proteins using enhanced phylogenetic trees.

MOTIVATION: Functional linkages implicate pairwise relationships between proteins that work together to implement biological tasks. During evolution, functionally linked proteins are likely to be preserved or eliminated across a range of genomes in a correlated fashion. Based on this hypothesis, phylogenetic profiling-based approaches try to detect pairs of protein families that show similar evolutionary patterns. Traditionally, the evolutionary pattern of a protein is encoded by either a binary profile of presence and absence of this protein across species or an occurrence profile that indicates the distribution of copies of this protein across species. RESULTS: In our study, we characterize each protein by its enhanced phylogenetic tree, a novel graphical model of the evolution of a protein family with explicitly marked by speciation and duplication events. By topological comparison between enhanced phylogenetic trees, we are able to detect the functionally associated protein pairs. Because the enhanced phylogenetic trees contain more evolutionary information of proteins, our method shows greater performance and discovers functional linkages among proteins more reliably compared with the conventional approaches.

Inferring the physical connectivity of complex networks from their functional dynamics.

BACKGROUND: Biological networks, such as protein-protein interactions, metabolic, signalling, transcription-regulatory networks and neural synapses, are representations of large-scale dynamic systems. The relationship between the network structure and functions remains one of the central problems in current multidisciplinary research. Significant progress has been made toward understanding the implication of topological features for the network dynamics and functions, especially in biological networks. Given observations of a network system's behaviours or measurements of its functional dynamics, what can we conclude of the details of physical connectivity of the underlying structure? RESULTS: We modelled the network system by employing a scale-free network of coupled phase oscillators. Pairwise phase coherence (PPC) was calculated for all the pairs of oscillators to present functional dynamics induced by the system. At the regime of global incoherence, we observed a Significant pairwise synchronization only between two nodes that are physically connected. Right after the onset of global synchronization, disconnected nodes begin to oscillate in a correlated fashion and the PPC of two nodes, either connected or disconnected, depends on their degrees.Based on the observation of PPCs, we built a weighted network of synchronization (WNS), an all-to-all functionally connected network where each link is weighted by the PPC of two oscillators at the ends of the link. In the regime of strong coupling, we observed a Significant similarity in the organization of WNSs induced by systems sharing the same substrate network but different configurations of initial phases and intrinsic frequencies of oscillators.We reconstruct physical network from the WNS by choosing the links whose weights are higher than a given threshold. We observed an optimal reconstruction just before the onset of global synchronization.Finally, we correlated the topology of the background network to the observed change of the functional activities in the system. CONCLUSIONS: The results presented in this study indicate a strong relationship between the structure and dynamics of complex network systems. As coupling strength increases, synchronization emerges among hub nodes and recruits small-degree nodes. The results show that the onset of global synchronization in the system hinders the reconstruction of an underlying complex structure. Our analysis helps to clarify how the synchronization is achieved in systems of different network topologies.

Evaluation of different domain-based methods in protein interaction prediction.

Protein-protein interactions (PPIs) play an important role in many biological functions. PPIs typically involve binding between domains, the basic units of protein folding, evolution and function. Identifying domain-domain interactions (DDIs) would aid understanding PPI networks. Recently, many computational methods aimed to infer DDIs from databases of interacting proteins and subsequently used the inferred DDIs to predict new PPIs. We attempt to describe systematically current domain-based approaches including the association method, maximum likelihood estimation and parsimonious explanation method. The performance of these methods at inferring DDIs and predicting PPIs was evaluated comparatively. We observe that each method generates artefacts in certain situations and discuss biases in the available benchmark sets.