Periodic Reporting for period 1 - OrganoidAlign (Comparative modelling of organoids in the age of single-cell transcriptomics)
Reporting period: 2021-06-15 to 2023-06-14
The OrganoidAlign project was initiated to develop a statistically-consistent and robust computational framework for comparing an in vitro organoid system to its in vivo reference tissue using their single cell transcriptomic (scRNA-seq) profiles. This is a fully data-driven interdisciplinary research project, where both computational and immunology expertise came together to design and develop relevant computational methods for comparative modelling of organoids at single-cell resolution.
Organoids are three-dimensional self-assembling constructs of tissue that are grown in vitro by differentiating either from induced pluripotent stem cells (iPSCs) or primary adult stem cells. As proxies to their in vivo counterparts, organoids can act as model systems enabling mechanistic studies in vitro to understand human developmental biology, as well as to drive targeted cell engineering and regenerative medicine. Thus it is important to develop organoids that are more closely and accurately following the dynamics of their in vivo tissue reference, being faithful and reliable. As a rapidly advancing field, this has created a timely need for computational methods that can quantitatively assess the degree of recapitulation of an organoid compared to its in vivo reference, and infer matches and mismatches between them in terms of their expected cell type/state compositions and transcriptomics across their developmental time.
This proposal aimed to develop computational methods to measure the reliability of an organoid based on its scRNA-seq profile under a statistical and information theoretic framework. This particularly addressed the computational problem of trajectory alignment. The main objective was to develop a Bayesian information-theoretic alignment framework that can capture the matches and mismatches between an organoid system and its in vivo tissue reference, using single-cell RNA sequencing data. Our main organoid system of focus was an in-house grown artificial thymus organoid (ATO) system which mimics T cell differentiation from iPSCs to mature Single Positive T cells.
Organoids are three-dimensional self-assembling constructs of tissue that are grown in vitro by differentiating either from induced pluripotent stem cells (iPSCs) or primary adult stem cells. As proxies to their in vivo counterparts, organoids can act as model systems enabling mechanistic studies in vitro to understand human developmental biology, as well as to drive targeted cell engineering and regenerative medicine. Thus it is important to develop organoids that are more closely and accurately following the dynamics of their in vivo tissue reference, being faithful and reliable. As a rapidly advancing field, this has created a timely need for computational methods that can quantitatively assess the degree of recapitulation of an organoid compared to its in vivo reference, and infer matches and mismatches between them in terms of their expected cell type/state compositions and transcriptomics across their developmental time.
This proposal aimed to develop computational methods to measure the reliability of an organoid based on its scRNA-seq profile under a statistical and information theoretic framework. This particularly addressed the computational problem of trajectory alignment. The main objective was to develop a Bayesian information-theoretic alignment framework that can capture the matches and mismatches between an organoid system and its in vivo tissue reference, using single-cell RNA sequencing data. Our main organoid system of focus was an in-house grown artificial thymus organoid (ATO) system which mimics T cell differentiation from iPSCs to mature Single Positive T cells.
This project successully achieved all the main scientific objectives and training objectives which were defined in the proposal.
Scientific objectives:
1. Understand and define the scRNA-seq profile comparison problem under the Minimum Message Length (MML) criterion, through fully parameterized models.
2. Formulate and develop the framework with defined inference components and a statistically robust alignment quality measure.
3. Test and validate the framework’s applicability as a guide for informing variations and missing components between the in vitro and in vivo models.
Training objectives:
1. Knowledge acquisition.
2. Career development and improvement of scientific communication, presentation, and networking skills.
3. Improvement of project management and planning skills.
4. Knowledge transfer and outreach.
The main result of this project is a new state of the art, single-cell trajectory alignment framework (named Genes2Genes), which is now available as an open source, Python pacakge with documentation at: https://github.com/Teichlab/Genes2Genes(opens in new window).
The main manuscript of the study is currently under revision, and available as a preprint at bioRxiv (DOI: https://doi.org/10.1101/2023.03.08.531713(opens in new window)).
This study investigated the similarities and differences of T cell differentiation between our in-house grown artificial thymus organoid (ATO) system and a pan fetal reference.
We found that the last stage of Single Positive T cell maturity differs between the in vitro and in vivo systems in terms of the transcriptional factors (TFs), where most of the distant TFs are associated with the TNFa singaling via NFkB pathway.
Scientific objectives:
1. Understand and define the scRNA-seq profile comparison problem under the Minimum Message Length (MML) criterion, through fully parameterized models.
2. Formulate and develop the framework with defined inference components and a statistically robust alignment quality measure.
3. Test and validate the framework’s applicability as a guide for informing variations and missing components between the in vitro and in vivo models.
Training objectives:
1. Knowledge acquisition.
2. Career development and improvement of scientific communication, presentation, and networking skills.
3. Improvement of project management and planning skills.
4. Knowledge transfer and outreach.
The main result of this project is a new state of the art, single-cell trajectory alignment framework (named Genes2Genes), which is now available as an open source, Python pacakge with documentation at: https://github.com/Teichlab/Genes2Genes(opens in new window).
The main manuscript of the study is currently under revision, and available as a preprint at bioRxiv (DOI: https://doi.org/10.1101/2023.03.08.531713(opens in new window)).
This study investigated the similarities and differences of T cell differentiation between our in-house grown artificial thymus organoid (ATO) system and a pan fetal reference.
We found that the last stage of Single Positive T cell maturity differs between the in vitro and in vivo systems in terms of the transcriptional factors (TFs), where most of the distant TFs are associated with the TNFa singaling via NFkB pathway.
The new alignment framework Genes2Genes developed under this project addresses the limitations in the current state of the art trajectory alignment approaches.
It facilitates a statistically robust and consistent alignment of two single-cell trajectories at both gene-level and cell-level, enabling users to apply this framework not only to in vitro and in vivo comparisons, but also for many other general contexts such as healthy versus disease trajectories, control versus treatment trajectories, cross species comparison and so on.
The in vitro versus in vivo comparison we had performed demonstrates the power of computational alignment, especially its usecase in facilitating the refinement of an organoid protocol to achieve more accuracy in its ability to recapitulate in vivo dynamics. Overall, the outcomes of this project have direct benefits towards developmental biology and regenerative medicine, which in turn help the scientists to improve human health across the globe.
It facilitates a statistically robust and consistent alignment of two single-cell trajectories at both gene-level and cell-level, enabling users to apply this framework not only to in vitro and in vivo comparisons, but also for many other general contexts such as healthy versus disease trajectories, control versus treatment trajectories, cross species comparison and so on.
The in vitro versus in vivo comparison we had performed demonstrates the power of computational alignment, especially its usecase in facilitating the refinement of an organoid protocol to achieve more accuracy in its ability to recapitulate in vivo dynamics. Overall, the outcomes of this project have direct benefits towards developmental biology and regenerative medicine, which in turn help the scientists to improve human health across the globe.