Periodic Reporting for period 1 - MOLiNTEL (Molecular Basis of Intelligence)
Periodo di rendicontazione: 2021-07-01 al 2023-06-30
In the MOLiNTEL project, we attempt to understand how neurons communicate with each other to respond to external stimuli at the very molecular level. To address this complex and multifaceted problem we focused on analyzing the underlying signaling machinery of the neuron-neuron junction called the “synapse”. The most crucial components of this machinery are proteins from the iGluR family. iGluRs are responsible for making neurons “plastic” i.e. they allow neurons either strengthen or weaken connection based on the type and the volume of signaling between the neurons. This neuronal plasticity is the crux of our ability to learn, memorize and infer knowledge. Failures in the iGluR machinery are directly connected to diseases associated with neurodegenerative diseases such as Alzheimer’s. To develop medication specifically targeting the defective functionality of these proteins it is of great importance to understand the complex dynamics of these proteins which are responsible for the failure states. Using state-of-the-art molecular modelling and high throughput artificial intelligence methods we wanted to unravel the relation between the structure, dynamics, and function of the members of the iGluR family. Once these relationships are pinned down, then it is possible to concentrate on the task of building defect-specific therapeutic approaches to solve medical issues. To identify this relationship, we also aimed to build a computational machinery that can systematically and objectively explore the dynamic landscape of the iGluR family members in an efficient manner so that we could connect the function of these proteins to their structure. This was built in the form of a software framework operating on top of and posterior to all available simulation protocols. The advantage of this framework is of course more general, and it can be used in alternative scenarios involving other simulation systems. At the conclusion of our two years of work, we built the software framework. Using this framework, we were successfully able to find the mechanism of activation of two members of iGluR family, named AMPA-receptors and NMDA-receptors. In addition, the software tools developed as a part of the project have allowed us to investigate other protein families associated with neuronal signaling, such as the GPCR family. We are currently comparing the two mechanisms to each other to try and understand how and why these mechanisms differ between the members of the iGluR family. Such a comparison is meant to give us insight into how these members can be differentially targeted by pharmaceutical approaches despite their structural similarities.
As a part of the MOLiNTEL project, we have achieved three concrete scientific results.
First, we have built a software framework called Function Sampling Technique (FST) that is capable of efficiently sampling the relevant dynamics and states of the protein corresponding to the function of the protein are explored. The engine behind the software uses Artificial Intelligence (AI) and Machine Learning (ML) to explore the structural landscape of the protein and encourages the discovery of hitherto unencountered states specifically so that they are along paths that explore the function of the protein. These states are then enriched so that a robust model of the structure-function relationship can be built. This software has been currently released through a publicly available repository on the platform GitHub to be shared in open access. We are currently in the process of submitting the publication corresponding to its implementation. For the MONiNTEL project, we have used FST to generate models to understand the activation mechanism of two members of the iGluR family, the NMDA-receptor, and the AMPA-receptor.
Second, using the FST software we have unraveled the mechanism of activation of the NMDA-Receptor. We have also discovered its active state structure, which was hitherto unknown in literature in a fully atomistic form. Using our ML and AI-based tools we have also discovered the structural mechanism through which it undergoes deactivation in the absence of neuronal signaling and the mechanism of its “desensitization” i.e. the behavior where the receptor’s signaling capacity is diminished in response to a prolonged state of activation. A research article combining these results is currently under preparation to be submitted soon.
Third, we have developed a new methodology termed apples2apples which we are using to compare the dynamics of proteins from the same family (such as members of the iGluR family). In this approach, we use bioinformatics approaches combined with AI techniques to create a common descriptive framework to examine the dynamics of the proteins. This common framework allows us to discriminate if the proteins have the same or differing mechanisms that allow them to perform their function. This approach has been successfully implemented and the members of the iGluR family are under investigation to understand the origins of how two structurally similar proteins have differential behavior in their dynamics.
Aside from the scientific results, one of the key objectives of this MSCA fellowship was the training of the grant holder. One of the main stated goals in the application on that front was to train the grant holder as a scientific adviser. During the two-year span of the fellowship, two master's degree projects and one bachelor's thesis project have been completed on the topic associated with the grant. Both master's degree holders achieved the highest grade (5/5) in their submissions at the University of Helinski. Also, the applicant has been involved as a co-supervisor in a PhD thesis to be submitted this year.
First, we have built a software framework called Function Sampling Technique (FST) that is capable of efficiently sampling the relevant dynamics and states of the protein corresponding to the function of the protein are explored. The engine behind the software uses Artificial Intelligence (AI) and Machine Learning (ML) to explore the structural landscape of the protein and encourages the discovery of hitherto unencountered states specifically so that they are along paths that explore the function of the protein. These states are then enriched so that a robust model of the structure-function relationship can be built. This software has been currently released through a publicly available repository on the platform GitHub to be shared in open access. We are currently in the process of submitting the publication corresponding to its implementation. For the MONiNTEL project, we have used FST to generate models to understand the activation mechanism of two members of the iGluR family, the NMDA-receptor, and the AMPA-receptor.
Second, using the FST software we have unraveled the mechanism of activation of the NMDA-Receptor. We have also discovered its active state structure, which was hitherto unknown in literature in a fully atomistic form. Using our ML and AI-based tools we have also discovered the structural mechanism through which it undergoes deactivation in the absence of neuronal signaling and the mechanism of its “desensitization” i.e. the behavior where the receptor’s signaling capacity is diminished in response to a prolonged state of activation. A research article combining these results is currently under preparation to be submitted soon.
Third, we have developed a new methodology termed apples2apples which we are using to compare the dynamics of proteins from the same family (such as members of the iGluR family). In this approach, we use bioinformatics approaches combined with AI techniques to create a common descriptive framework to examine the dynamics of the proteins. This common framework allows us to discriminate if the proteins have the same or differing mechanisms that allow them to perform their function. This approach has been successfully implemented and the members of the iGluR family are under investigation to understand the origins of how two structurally similar proteins have differential behavior in their dynamics.
Aside from the scientific results, one of the key objectives of this MSCA fellowship was the training of the grant holder. One of the main stated goals in the application on that front was to train the grant holder as a scientific adviser. During the two-year span of the fellowship, two master's degree projects and one bachelor's thesis project have been completed on the topic associated with the grant. Both master's degree holders achieved the highest grade (5/5) in their submissions at the University of Helinski. Also, the applicant has been involved as a co-supervisor in a PhD thesis to be submitted this year.
The MOLiNTEL project has pushed the boundaries of simulation techniques used to computationally investigate protein dynamics and kinetics further while using these advances to understand the activation mechanism of members of the iGluR family. The methodologies developed as a part of the project are highly transferable towards challenges encountered often in the field of biomolecular simulations. During this project, we have demonstrated that these methodologies can be directly used to understand unrelated families of proteins such as GPCRs with great success. In addition to the purely methodological advances, we have uncovered the hitherto unknown active state of the NMDA-receptor in the iGluR family and the connection of its functional domain to the dynamics of structurally distal region of the protein. This discovery is of great potential pharmaceutical interest as it can pinpoint how extant drugs modulate the iGluR family members. At the same time, the results of the project have created a path for the future development of drugs that can be more effective than the current drugs leading to novel therapeutic strategies. The most significant impact of the results of MOLiNTEL is gaining a molecular-level understanding of how the machinery behind human intelligence works. Our work has shed light on important aspects of the questions about how the signal from one neuron reaches another. While answering these questions we have also made significant progress to develop strategies to combat neurological diseases such as Alzheimer’s disease which is a leading problem in the ageing population of the first world.