Search for: All records

In the field of materials science, microscopy is the first and often only accessible method for structural characterization. There is a growing interest in the development of machine learning methods that can automate the analysis and interpretation of microscopy images. Typically training of machine learning models requires large numbers of images with associated structural labels, however, manual labeling of images requires domain knowledge and is prone to human error and subjectivity. To overcome these limitations, we present a semi-supervised transfer learning approach that uses a small number of labeled microscopy images for training and performs as effectively as methods trained on significantly larger image datasets. Specifically, we train an image encoder with unlabeled images using self-supervised learning methods and use that encoder for transfer learning of different downstream image tasks (classification and segmentation) with a minimal number of labeled images for training. We test the transfer learning ability of two self-supervised learning methods: SimCLR and Barlow-Twins on transmission electron microscopy (TEM) images. We demonstrate in detail how this machine learning workflow applied to TEM images of protein nanowires enables automated classification of nanowire morphologies ( e.g. , single nanowires, nanowire bundles, phase separated) as well as segmentation tasks that can serve as groundwork for quantification of nanowire domain sizes and shape analysis. We also extend the application of the machine learning workflow to classification of nanoparticle morphologies and identification of different type of viruses from TEM images. 

Collagen-targeting strategies have proven to be an effective method for targeting drugs to pathological tissues for treatment of disease. The use of collagen-like peptides for controlling the assembly of drug delivery vehicles, as well as their integration into collagen-containing matrices, offers significant advantages for tuning the morphologies of assembled structures, their thermoresponsiveness, and the loading and release of both small-molecule and macro-molecular cargo. In this contribution, we summarize the design and development of collagen-peptide-based drug delivery systems introduced by the Kiick group and detail the expansion of our understanding and the application of these unique molecules through collaborations with experts in computational simulations (Jayaraman), osteoarthritis (Price), and gene delivery (Sullivan). Kiick was inducted as a Fellow of the National Academy of Inventors in 2019 and was to deliver an address describing the innovations of her research. Given the cancellation of the NAI Annual Meeting as a result of coronavirus travel restrictions, her work based on collagen-peptide-mediated assembly is instead summarized in this contribution. 

Assembling peptides allow the creation of structurally complex materials, where amino acid selection influences resulting properties. We present a synergistic approach of experiments and simulations for examining the influence of natural and non-natural amino acid substitutions via incorporation of charged residues and a reactive handle on the thermal stability and assembly of multifunctional collagen mimetic peptides (CMPs). Experimentally, we observed inclusion of charged residues significantly decreased the melting temperature of CMP triple helices with further destabilization upon inclusion of the reactive handle. Atomistic simulations of a single CMP triple helix in explicit water showed increased residue-level and helical structural fluctuations caused by the inclusion of the reactive handle; however, these atomistic simulations cannot be used to predict changes in CMP melting transition. Coarse-grained (CG) simulations of CMPs at experimentally relevant solution conditions, showed, qualitatively, the same trends as experiments in CMP melting transition temperature with CMP design. These simulations show that when charged residues are included electrostatic repulsions significantly destabilize the CMP triple helix and that an additional inclusion of a reactive handle does not significantly change the melting transition. Based on findings from both experiments and simulations, the sequence design was refined for increased CMP triple helix thermal stability, and the reactive handle was utilized for the incorporation of the assembled CMPs within covalently crosslinked hydrogels. Overall, a unique approach was established for predicting stability of CMP triple helices for various sequences prior to synthesis, providing molecular insights for sequence design towards the creation of bulk nanostructured soft biomaterials. 

Peptide–polymer conjugates are a class of soft materials composed of covalently linked blocks of protein/polypeptides and synthetic/natural polymers. These materials are practically useful in biological applications, such as drug delivery, DNA/gene delivery, and antimicrobial coatings, as well as nonbiological applications, such as electronics, separations, optics, and sensing. Given their broad applicability, there is motivation to understand the molecular and macroscale structure, dynamics, and thermodynamic behavior exhibited by such materials. We focus on the past and ongoing molecular simulation studies aimed at obtaining such fundamental understanding and predicting molecular design rules for the target function. We describe briefly the experimental work in this field that validates or motivates these computational studies. We also describe the various models (e.g., atomistic, coarse-grained, or hybrid) and simulation methods (e.g., stochastic versus deterministic, enhanced sampling) that have been used and the types of questions that have been answered using these computational approaches. 

Guest Editors Arthi Jayaraman and Amish Patel introduce this themed collection of papers showcasing the latest research on the molecular design and engineering of bioinspired, biological and/or biomimetic materials.