A Machine Learning-Based Investigation of Integrin Expression Patterns in Cancer and Metastasis.

Background: Integrins, a family of transmembrane receptor proteins, play complex roles in cancer development and metastasis. These roles could be better delineated through machine learning of transcriptomic data to reveal relationships between integrin expression patterns and cancer. Methods: We collected publicly available RNA-Seq integrin expression from 8 healthy tissues and their corresponding tumors, along with data from metastatic breast cancer. We then used machine learning methods, including t-SNE visualization and Random Forest classification, to investigate changes in integrin expression patterns. Results: Integrin expression varied across tissues and cancers, and between healthy and cancer samples from the same tissue, enabling the creation of models that classify samples by tissue or disease status. The integrins whose expression was important to these classifiers were identified. For example, ITGA7 was key to classification of breast samples by disease status. Analysis in breast tissue revealed that cancer rewires co-expression for most integrins, but the co-expression relationships of some integrins remain unchanged in healthy and cancer samples. Integrin expression in primary breast tumors differed from their metastases, with liver metastasis notably having reduced expression. Conclusions: Integrin expression patterns vary widely across tissues and are greatly impacted by cancer. Machine learning of these patterns can effectively distinguish samples by tissue or disease status.

Map conformational landscapes of intrinsically disordered proteins with polymer physics quantities.

Disordered proteins are conformationally flexible proteins that are biologically important and have been implicated in devastating diseases such as Alzheimer's disease and cancer. Unlike stably folded structured proteins, disordered proteins sample a range of different conformations that needs to be accounted for. Here, we treat disordered proteins as polymer chains, and compute a dimensionless quantity called instantaneous shape ratio (Rs), as Rs = Ree2/Rg2, where Ree is end-to-end distance and Rg is radius of gyration. Extended protein conformations tend to have high Ree compared with Rg, and thus have high Rs values, whereas compact conformations have smaller Rs values. We use a scatter plot of Rs (representing shape) against Rg (representing size) as a simple map of conformational landscapes. We first examine the conformational landscape of simple polymer models such as Random Walk, Self-Avoiding Walk, and Gaussian Walk (GW), and we notice that all protein/polymer maps lie within the boundaries of the GW map. We thus use the GW map as a reference and, to assess conformational diversity, we compute the fraction of the GW conformations (fC) covered by each protein/polymer. Disordered proteins all have high fC scores, consistent with their disordered nature. Each disordered protein accesses a different region of the reference map, revealing differences in their conformational ensembles. We additionally examine the conformational maps of the nonviral gene delivery vector polyethyleneimine at various protonation states, and find that they resemble disordered proteins, with coverage of the reference map decreasing with increasing protonation state, indicating decreasing conformational diversity. We propose that our method of combining Rs and Rg in a scatter plot generates a simple, meaningful map of the conformational landscape of a disordered protein, which in turn can be used to assess conformational diversity of disordered proteins.

Exploring Structures and Dynamics of Protamine Molecules through Molecular Dynamics Simulations.

Protamines are arginine-rich proteins that condense DNA in sperm. Despite their importance in reproduction, information on protamine structure is scarce. We, therefore, used molecular dynamics to examine the structures of salmon, bull P1, and human P1 protamines. The sizes and shapes of each protamine varied widely, indicating that they were disordered with structures covering a broad conformational landscape, from hairpin loop structures to extended coils. Despite their general disorder, the protamines did form secondary structures, including helices and hairpin loops. In eutherians, hairpins may promote disulfide bonding that facilitates protamine-DNA condensation, but the specifics of this bonding is not well established. We examined inter-residue distances in the simulations to predict residue pairs likely to form intramolecular bonds, leading to the identification of bonding pairs consistent with previous results in bull and human. These results support a model for eutherian protamine structures where a highly charged center is surrounded by disulfide-bond-stabilized loops.