A broad and simple definition of `language' is a set of sequences constructed from a finite set of symbols. By this definition, biological sequences, human languages, and many sequential phenomena that exist in the world can be viewed as languages. Although this definition is simple, it includes languages employing very complicated grammars in the creation of their sequences of symbols. Examples are biophysical principles governing biological sequences (e.g., DNA, RNA, and protein sequences), as well as grammars of human languages determining the structure of clauses and sentences. This dissertation uses a language-agnostic point of view in the processing of both biological sequences and human languages. Two main strategies are adopted toward this purpose, (i) character-level, or more accurately, subsequence-level processing of languages, which allows for simple modeling of the sequence similarities based on local information or, bag-of-subsequences, (ii) language model based representation learning encoding contextual information of sequence elements using the neural network language models. I propose language-agnostic and subsequence-based language processing using the above-mentioned strategies in addressing three main research problems in proteomics, genomics/metagenomics, and natural languages using the same point-of-view.

One of the main challenges in proteomics is that there exists a large gap between the number of known protein sequences and known protein structures/functions. The central question here is how to efficiently use large numbers of sequences to achieve a better performance in the structural and functional annotation of protein sequences. Here, we proposed subsequence-based representations of protein sequences and their language model-based embeddings trained over a large dataset of protein sequences, which we called protein vectors (or ProtVec). In addition, we introduced a motif discovery approach, benefiting from probabilistic segmentation of protein sequences to find functional and structural motifs. This segmentation is also inferred from large protein sequence datasets. The ProtVec approach has proved a seminal contribution in protein informatics and now is widely used for machine learning based protein structure and function annotations. We showed in different protein informatics tasks that bag-of-subsequences and protein embeddings are complementary information for language-agnostic prediction of protein structures and functions, which also achieved the state-of-the-art performance in the 2 out of 3 tasks of Critical Assessment of protein Function Annotation (CAFA) in 2018 (CAFA 3.14). Moreover, we systematically investigated the role of representation and deep learning architecture in protein secondary structure prediction from the primary sequence. Publicly available tools are provided for achieving state-of-the-art performance accuracy that can be further expanded by the community.

One of the prominent challenges in metagenomics involves the host phenotypic characterization based on the associated microbial samples. Microbial communities exist almost on every accessible surface on earth, supporting, regulating, and even causing unwanted conditions (e.g., diseases) to their hosts and environments. Detection of the host phenotype and the phenotype-specific taxa from the microbial samples is the chief goal here. For instance, identifying distinctive taxa for microbiome-related diseases is considered key to the establishment of diagnosis and therapy options in precision medicine and imposes high demands on the accuracy of microbiome analysis techniques. Here, we propose two distinct language-agnostic subsequence-based processing methods for machine learning on 16S rRNA sequencing, currently the most cost-effective approach for sequencing of microbial communities. We propose alignment- and reference- free methods, called MicroPheno and DiTaxa, designed for microbial phenotype and biomarker detection, respectively. MicroPheno is a k-mer based approach achieving the state-of-the-art performance in the host phenotype prediction from 16S rRNA outperforming conventional OTU features. DiTaxa, substitutes standard OTU-clustering by segmenting 16S rRNA reads into the most frequent variable-length subsequences. We compared the performance of DiTaxa to the state-of-the-art methods in phenotype and biomarker detection, using human-associated 16S rRNA samples for periodontal disease, rheumatoid arthritis, and inflammatory bowel diseases, as well as a synthetic benchmark dataset. DiTaxa performed competitively to MicroPheno (state-of-the-art approach) in phenotype prediction while outperforming the OTU-based state-of-the-art approach in finding biomarkers in both resolution and coverage evaluated over known links from literature and synthetic benchmark datasets.

The third central problem we addressed in this dissertation is focused on human languages. Many of 7000 world's natural languages are low-resource and lack digitized linguistic resources. This has put many of these human languages in danger of extinction and has motivated developing methods for automatic creation of linguistic resources and linguistic knowledge for low-resource languages. To address this problem via our language-agnostic point of view (by not treating different languages differently), we develop SuperPivot for subsequence-based linguist marker detection in parallel corpora of 1000 languages, which was the first computational investigation for linguistic resource creation in such a scale. As an example, SuperPivot was used to study the typology of tense in 1000 languages. Next, we utilized SuperPivot for the creation of the largest sentiment lexicon to date in terms of the number of covered languages (1000+ languages) achieving macro-F1 over 0.75 on word sentiment prediction for most evaluated languages, meaning that we enable sentiment analysis in many low resource languages. To ensure the usability of UniSent lexica for any new domain, we propose DomDrift, a method quantifying the semantic changes of words in the sentiment lexicon in the new domain. Next, we extend the DomDrift method to quantifying the semantic changes of all words in the language. We proposed a new metric for language comparisons based on the language word embedding graphs requiring only monolingual embeddings and word mapping between languages obtained through statistical alignment in parallel corpora. We performed language comparison for fifty natural languages and twelve genetic language variations of different organisms. As a result, natural languages of the same family were clustered together. In addition, applying the same method on organisms' genomes confirmed a high-level difference in the genetic language model of humans/animals versus plants. This method called word embedding language divergence is a step toward unsupervised or minimally supervised comparison of languages in their broad definition.

Focal adhesions are critical to cellular processes such as cell migration and cell-substrate adhesion. Focal adhesion formation can be mechanically regulated: forces either from outside the cell or from contracting actin filaments can induce rapid growth and maturation of the focal adhesions. One hypothesis explored here contends that force-induced focal adhesion formation results from mechanosensing by individual protein components. Molecular dynamics computational simulations are developed to evaluate mechanosensing by talin and vinculin. In Part 1 of this dissertation, two force-induced conformational changes are suggested for talin's activation: (i) the cryptic vinculin-binding sites (VBS) can be activated by stretch of an individual talin rod domain, and (ii) talin can adopt multiple dimer orientations in response to forces applied from outside the cell. The mechanisms of vinculin activation are then explored in Part 2. The trajectory of a vinculin conformational changes that would render it activated is predicted along with the structure of an activated vinculin. Domain 1 (D1) is predicted to separate from the vinculin tail (Vt) during activation. In this context, the PIP2 from the cell membrane is shown to preferentially bind basic residues on the vinculin surface and recruit vinculin to membrane proximal regions, potentially allowing for vinculin phosphorylation. The impact of phosphorylation on the vinculin structure is simulated and it is suggested that phosphorylation could prime vinculin for activation by reducing the strength of inter-domain interactions stabilizing the auto-inhibited vinculin conformation. In Part 3 of this dissertation, the interaction of activated vinculin with its binding partners is simulated. It is demonstrated that a talin VBS can only link vinculin prior to activation but can completely bind vinculin following activation. Vinculin activation by movement at D1 is shown to be necessary and sufficient for linking vinculin to actin filaments. Furthermore, simulation of F-actin caping by vinculin suggests that a second vinculin conformational change, releasing Vt from all vinculin head domains, facilitates effective capping of the actin filament. Three vinculinbinding sites on F-actin are predicted. These simulations demonstrate that indeed both talin and vinculin can exhibit molecular mechanosensing.

Physical environment guides tissue regeneration and morphology in both health and disease. In the past three decades, several experiments illustrated that mechanical cues are captured and transduced to biochemical signals in the cellular level (mechanotransduction) mediated by cell adhesion. Cells adhere to their microenvironment through large protein assemblies known as focal adhesions that directly couple intra- and extra-cellular matrices and play a critical role in many vital cell functions including proliferation, differentiation and cell fate. It is inherently difficult to investigate the molecular basis of focal adhesion formation and growth using current experimental methodologies due to the fine time- and length-scales of protein-protein interactions. Here, I used molecular dynamics simulations to investigate the underlying molecular mechanisms of focal adhesion formation and maturation with atomic resolution.

Integrins are key focal adhesion receptors that reside on the cell membrane and mediate bi-directional signaling between cell cytoskeleton and ECM. Focal adhesions are a mixture of integrin-associated protein complexes known as integrin modules that forms the basic adhesion units. Integrin module formation is initiated by talin binding to the integrin tail, which is shown to be sufficient for integrin activation. A few other focal adhesion proteins can also directly engage with integrin’s cytoplasmic tail and link it to the actin cytoskeleton. It is not yet clear how simultaneous (cooperative) versus sequential (competitive) binding of focal adhesion proteins to integrin with respect to talin result in different functionalities of integrin modules. In the first part of this study, competitive versus cooperative integrin binding between two important focal adhesion proteins –filamin and α-actinin– with talin were studied. The purpose of this aim was to gain insight on integrin module formation that eventually determines its functional properties.

Maturation of focal adhesions follows an increase in local forces. A well-established hypothesis on force transmission across focal adhesion complexes is the presence of mechanosensitive elements that change their conformations in response to force. In the second part of this study, we investigated and argued force-induced conformational changes of two important focal adhesion proteins –vinculin and α-actinin– in order to shed light on their role in transmission of forces across focal adhesions leading to adhesion maturation and growth.

In conclusion, this study unravels the structural basis of mechanosensitivity of key focal adhesions. Furthermore, important molecular interactions that give rise to mechanosensitivite characteristics of focal adhesions were studies. Important impacts of the current study include but are not limited to the following: 1) our results was used to complement previous experimental studies and also construct new hypotheses for future experiments; 2) Understanding regulatory mechanisms of focal adhesions is critical for developing novel therapeutics for many diseases involving cell adhesion including cancer as it enhances target recognition and the accuracy of drug delivery systems. 3) In addition, performing extensive simulations on protein complexes will contribute to improving the accuracy of various aspects of computational methods including empirical force fields that is indicative of our understanding of fundamental physical and chemical principles governing protein-protein interactions. 4) And most importantly, this work provides a fundamental insight into the relation between structure and function of mechanosensitive proteins in focal adhesions.

The dynamic nature of cellular attachment and detachment which regulates cell function must be characterized to understand essential biological phenomena such as embryonic development, inflammatory immune response, wound repair, cancer metastasis, etc. Although extensive efforts have been devoted toward cell motility manipulation for a better understanding cellular dynamics, these efforts have remained in their infancy. A programmable manipulation of adherent-cell motility therefore has remained a challenging task. The goal of this research is to design addressable, multi-functional, and reusable biological platforms, nicknamed biological breadboards (BBBs), to enable the spatiotemporal manipulation of cell motility at cellular and even subcellular levels, offering a new method to analyze cellular dynamics in a quantitative way. In this thesis, this method for cell motility manipulation using BBBs and their applications to quantitative characterizations of single-cellular and intracellular dynamics are summarized.

The BBBs consisting of M×N gold electrodes, identical and independently operated, on Pyrex glass substrate are designed to secure their addressability, thereby achieving a high-degree-of-freedom in programmable manipulation of adherent cell motility. This programmable manipulation of adherent cell motility is implicated by two fundamental mechanisms of cell attachment and detachment. An arginine-glycine-aspartic (RGD) acid-terminated thiol is introduced to achieve those mechanisms by tethering the RGD to the gold electrodes via thiol linker. The first mechanism is solved by a RGD which is a recognition sequence for integrin binding to many extracellular matrix proteins. The RGD-terminated thiol connects a RGD peptide into the gold electrodes using its thiol end which shows a spontaneous chemisorption of R-S-H + Au -> R-S-AU + 1/2H2 where R is a substituent, thereby enhancing cell adhesion on the BBB. The second mechanism depends on a rapid reductive desorption of the thiol-gold self-assembled monolayer under negative bias voltage, R-S-Au + H+ + e- -> R-S-H + Au. This reaction makes the RGD-terminated thiol and even a cell or a part of the cell thereon detached from the gold electrodes. The BBBs are fabricated with microfabrication processes (optical lithography, e-beam evaporation, and lift-off) and their surfaces are modified by two surface-treatments of polyethylene glycol treatment on Pyrex glass and RGD-terminated thiol treatment on gold electrodes. The first is intended to achieve a cell-resistive surface which suppresses cell spreading on them, whereas the second is planned to tether a RGD peptide into gold via thiol linker, thereby providing a cell-adhesive surface.

In the experimental studies with a mouse embryonic fibroblast cell line (NIH 3T3), cellular and subcellular detachments have been made with the fabricated BBBs both to quantitatively characterize cellular and subcellular detachment dynamics and to investigate the addressability, multi-functionality, and reusability of the BBBs, thereby evaluating the BBBs as biological platforms for programmable manipulation of adherent cell motility. In addition, the changes in viscoelastic properties of the detached and retracting thin layer of a NIH 3T3 cell have been investigated.

The nuclear pore complex (NPC), located within the nuclear envelope of eukaryotic cells, acts as a pivotal conduit for the exchange of molecules between the nucleus and cytoplasm. The NPC functions as a selective gateway, permitting the free passage of small molecules (under 40 kDa) while restricting the movement of larger molecules, unless they are associated with nuclear transport receptors. This selectivity is essential for maintaining nuclear integrity as well as facilitating necessary material exchanges with the cytoplasm. The barrier function is regulated by phenylalanine- and glycine-rich nucleoporins (FG-Nups), which are intrinsically disordered proteins anchored to the NPC's inner surface. Despite a detailed understanding of FG-Nups' molecular compositions, the physical basis of their selective barrier formation remains poorly understood. Our study explored this issue through the lens of polymer physics theory and biomechanical simulations. We integrated field-theoretical and particle-based approaches to reveal that the selective transport mechanism is significantly influenced by the entropy variations of FG-Nups' conformations. We introduced a novel model that describes the molecular transport rate across the NPC, factoring in the dimensions of the pore, FG-Nups, and the molecules being transported. This research offers a new polymer physics-based framework to elucidate the complex selectivity of transport through the NPC.