Nylon 6 microfiltration membrane, for the treatment of brewery wastewater, was fabricated using an electrospinning technique, followed by hot-pressing. The fabricated membrane was robust and demonstrated highly hydrophilic property (water contact angle 39° at the touching point to the membrane surface and the water droplet was completely immersed into the membrane in 7 seconds), and higher porosity (65%) with pore sizes of 100 to 210 nm. The electrospun nylon 6 membrane showed higher pure water flux (850 LMH) at an applied pressure of 4 psi. The same membrane also demonstrated a 95% rejection rate of suspended solids (SS) in brewery wastewater treatment.

Japan's largest platform for academic e-journals: J-STAGE is a full text database for reviewed academic papers published by Japanese societies

Scope: This meta-analysis and review examine the conceptual and translational evolution of in silico methods in drug discovery, with emphasis on how chemical space, molecular similarity, cellular complexity, and localized singularities shape next-generation computational pharmacology. Building on Part 1, which defined the historical “what” and “when” of computational drug discovery, this Part 2 addresses the “how” and “why” by analyzing the principles that connect molecular descriptors, virtual screening, scaffold hopping, activity cliffs, and AI-driven lead prioritization. Methods: A systematic meta-analysis of 97 highly cited studies, selected from an initial pool of 300 top-cited publications, evaluates the ongoing paradigm shift from traditional computer-assisted drug discovery (CADD) toward advanced AI-driven workflows. Results: Thus, the current drug discovery practice enables the screening of ultra-large chemical libraries, mapping chemical diversity, generating hit maps, and prioritizing compounds through early ADMET assessment. The review shows that these methods can substantially accelerate early discovery and reduce experimental burden, but their impact remains mainly translational rather than automatically clinical. Limitations and Uncertainties: Major bottlenecks persist because model performance is constrained by chemical data quality, biological label uncertainty, missing or biased negative evidence, limited applicability-domain coverage, static representations of protein targets, incomplete treatment of solvent thermodynamics, and fragmented ADMET prediction. Conclusion: This work therefore argues that reliable AI-driven drug discovery requires more than faster prediction: it depends on fit-for-purpose datasets, explicit uncertainty handling, dynamic protein and solvent modeling, Systems ADMET, AI-PBPK integration, and closed-loop experimental validation. Overall, chemical space is presented not only as a repository of possible molecules, but as an operational framework for connecting molecular design, biological complexity, and clinically meaningful translation.