Evangelia Petsalaki
@epetsalaki.bsky.social
Group Leader at EMBL-EBI. Our group uses interdisciplinary approaches including network analysis, omics integration and mathematical modelling to study the context specific regulation of human cell signalling.
To make the SELPHI 2.0 probabilistic model widely useable for phosphoproteomics data interpretation we created a web server selphi2.com, where the user can provide their data and extract the kinases predicted to be regulating their phosphosites, extract context-specific networks and perform KSEA.
May 16, 2025 at 9:09 AM
To make the SELPHI 2.0 probabilistic model widely useable for phosphoproteomics data interpretation we created a web server selphi2.com, where the user can provide their data and extract the kinases predicted to be regulating their phosphosites, extract context-specific networks and perform KSEA.
Our resource largely expands upstream kinase annotations for phosphosites, provides distinct predictions even for paralog kinases and can help annotate understudied kinases using guilt-by-association. Moreover, we didn't find a difference in #of substrates between well studied and dark kinases.
May 16, 2025 at 9:09 AM
Our resource largely expands upstream kinase annotations for phosphosites, provides distinct predictions even for paralog kinases and can help annotate understudied kinases using guilt-by-association. Moreover, we didn't find a difference in #of substrates between well studied and dark kinases.
Overlapping our predictions with the two experimentally inferred kinase-substrate relationships we identify 76 high confidence kinase-substrate interactions, and fitting data onto our probabilistic model improves accuracy, providing a framework for extracting dataset-specific signalling networks.
May 16, 2025 at 9:09 AM
Overlapping our predictions with the two experimentally inferred kinase-substrate relationships we identify 76 high confidence kinase-substrate interactions, and fitting data onto our probabilistic model improves accuracy, providing a framework for extracting dataset-specific signalling networks.
Our method performs better than the state-of-the-art, in particular on unseen experimentally inferred kinase-substate relationships (from Hijazi et al and Sugiwama et al studies), and makes predictions for more kinases than any current method, including understudied ones.
May 16, 2025 at 9:09 AM
Our method performs better than the state-of-the-art, in particular on unseen experimentally inferred kinase-substate relationships (from Hijazi et al and Sugiwama et al studies), and makes predictions for more kinases than any current method, including understudied ones.
we found a signature of 24 genes accurtely predicting patient stage for our dataset and for two different public datasets. Our analysis showed which process deregulation is reflected in the plasma at which stage and we mapped approved drugs on these to be explored as potential stage-
January 19, 2025 at 11:18 AM
we found a signature of 24 genes accurtely predicting patient stage for our dataset and for two different public datasets. Our analysis showed which process deregulation is reflected in the plasma at which stage and we mapped approved drugs on these to be explored as potential stage-
Cell type deconvolution highlighted the changes in liver composition along the trajectory, with relative decreases of hepatocytes, increases in fibroblasts, cholangiocytes and immune cells. Associating these to the changes to the networks we identified cell type-specific deregulation networks.
January 19, 2025 at 11:18 AM
Cell type deconvolution highlighted the changes in liver composition along the trajectory, with relative decreases of hepatocytes, increases in fibroblasts, cholangiocytes and immune cells. Associating these to the changes to the networks we identified cell type-specific deregulation networks.
We then developed a graph-based approach to optimise the number of sliding windows along the trajectory and using network analysis we described the modulation dynamics of TFs and processes along the trajectory. e.g.
January 19, 2025 at 11:18 AM
We then developed a graph-based approach to optimise the number of sliding windows along the trajectory and using network analysis we described the modulation dynamics of TFs and processes along the trajectory. e.g.
We then used our published method from @charliegbarker.bsky.social (Barker et al, Genome Research, 2022) to create a reference MASLD network by associating gene co-expression modules with histology scores, identifying key transcription factors regulating them and applying network reconstruction.
January 19, 2025 at 11:18 AM
We then used our published method from @charliegbarker.bsky.social (Barker et al, Genome Research, 2022) to create a reference MASLD network by associating gene co-expression modules with histology scores, identifying key transcription factors regulating them and applying network reconstruction.
to place patients in the correct stage. We integrated transcriptomic data of MASLD patient liver biopsies. We used pseudotemporal ordering to place them along the disease continuum. We found 145 genes that also placed patients along the trajectory in independent datasets.
January 19, 2025 at 11:18 AM
to place patients in the correct stage. We integrated transcriptomic data of MASLD patient liver biopsies. We used pseudotemporal ordering to place them along the disease continuum. We found 145 genes that also placed patients along the trajectory in independent datasets.
Not sure what happened with that figure - here it is again!
December 10, 2024 at 9:58 AM
Not sure what happened with that figure - here it is again!
Finally, we found that the ARID1A KO cells had drastically reduced expression of MHCII molecules (through downregulation of RFX5, CIITA, RFXAP, RFXANK TFs, which we found also in ARID1A-mutant patient data) and changes in expression of ECM proteins, consistent with an immunosuppressive phenotype.
December 10, 2024 at 9:44 AM
Finally, we found that the ARID1A KO cells had drastically reduced expression of MHCII molecules (through downregulation of RFX5, CIITA, RFXAP, RFXANK TFs, which we found also in ARID1A-mutant patient data) and changes in expression of ECM proteins, consistent with an immunosuppressive phenotype.
We found PRKD1, FYN, JUN to be as key nodes where the two signals meet and using MaxFLow we found them to be key mediators of signalling from EGFR/ROS1 to JUN. PRKD1 and FYN are no longer upreg in ARID1A-KO cells, explaining the reactivation of JUN and resistance to treatment.
December 10, 2024 at 9:44 AM
We found PRKD1, FYN, JUN to be as key nodes where the two signals meet and using MaxFLow we found them to be key mediators of signalling from EGFR/ROS1 to JUN. PRKD1 and FYN are no longer upreg in ARID1A-KO cells, explaining the reactivation of JUN and resistance to treatment.
We then applied network propagation from the receptors that were different in the resistant state, and on those from the drug response to see where their effect meets. We repeated the propagation also from the EPH receptors which also had different kinase activities in ARID1A-KO.
December 10, 2024 at 9:44 AM
We then applied network propagation from the receptors that were different in the resistant state, and on those from the drug response to see where their effect meets. We repeated the propagation also from the EPH receptors which also had different kinase activities in ARID1A-KO.
The initial state of the ARID1A-KO cells was different, with several signalling proteins, including receptors being affected. To study how the ARID1A-KO initial state interferes with the drug response pathway we integrated the networks from factors 1(drug response) and factor 3 (genetic bg)
December 10, 2024 at 9:44 AM
The initial state of the ARID1A-KO cells was different, with several signalling proteins, including receptors being affected. To study how the ARID1A-KO initial state interferes with the drug response pathway we integrated the networks from factors 1(drug response) and factor 3 (genetic bg)
Looking into the differences between the drug response in sensitive vs resistant cells we surprisingly found that there was no difference at the omics/abundance level. However the kinase activities in the two cells where very different. How does that happen if they are using the exact same macinery?
December 10, 2024 at 9:44 AM
Looking into the differences between the drug response in sensitive vs resistant cells we surprisingly found that there was no difference at the omics/abundance level. However the kinase activities in the two cells where very different. How does that happen if they are using the exact same macinery?
Factor 2, associated with the combination treatment showed pathways related to DNA repair and Immune response to be downregulated and we also found several phosphosites related to EMT affected in the combination condition. Combination treatment didn't seem to kill the cells more.
December 10, 2024 at 9:44 AM
Factor 2, associated with the combination treatment showed pathways related to DNA repair and Immune response to be downregulated and we also found several phosphosites related to EMT affected in the combination condition. Combination treatment didn't seem to kill the cells more.
but we also found many changes in receptor tyrosine and other kinase activities, showing extensive signalling rewiring upon drug treatment. However we didn't observe any differences between single and combination treatment
December 10, 2024 at 9:44 AM
but we also found many changes in receptor tyrosine and other kinase activities, showing extensive signalling rewiring upon drug treatment. However we didn't observe any differences between single and combination treatment
We then dug into each factor to understand drug response in sensitive vs resistant ARID1A-KO cells. Factor 1, presented the drug response regardless of type or genetic background. We found (as previously known) strong downregulation of negative feedback regulators of the MAPK pathway,
December 10, 2024 at 9:44 AM
We then dug into each factor to understand drug response in sensitive vs resistant ARID1A-KO cells. Factor 1, presented the drug response regardless of type or genetic background. We found (as previously known) strong downregulation of negative feedback regulators of the MAPK pathway,
To study the mechanisms underpinning the variance in each of these factors we extracted signalling networks by adapting our tool phuEGO (Giudice et al, 2024) to use the weights from the factor analysis.
December 10, 2024 at 9:44 AM
To study the mechanisms underpinning the variance in each of these factors we extracted signalling networks by adapting our tool phuEGO (Giudice et al, 2024) to use the weights from the factor analysis.
We collected multiomics data upon single (BRAF/MEK) and combination drug treatment in A375 BRAFV600E cell lines and ARID1A-KO lines derived from these. Using MOFA we identified 3 factors associated with drug agnostic response, combination and genetic background.
December 10, 2024 at 9:44 AM
We collected multiomics data upon single (BRAF/MEK) and combination drug treatment in A375 BRAFV600E cell lines and ARID1A-KO lines derived from these. Using MOFA we identified 3 factors associated with drug agnostic response, combination and genetic background.
Resistance to BRAF/MAPK inhibitors is a significant challenge in melanoma treatment. We found ARID1A, which is mutated in >10% of all cancers with ambiguous roles in cancer growth and immune evasion, to confer resistance upon KO. We used it as a model to study mechanisms of melanoma drug resistance
December 10, 2024 at 9:44 AM
Resistance to BRAF/MAPK inhibitors is a significant challenge in melanoma treatment. We found ARID1A, which is mutated in >10% of all cancers with ambiguous roles in cancer growth and immune evasion, to confer resistance upon KO. We used it as a model to study mechanisms of melanoma drug resistance
Register now to join me at The Festival of Genomics & Biodata next month.
Registration is free for over 90% of the audience, click here to reserve your space: hubs.la/Q02TptWZ0
#FOG2025
Registration is free for over 90% of the audience, click here to reserve your space: hubs.la/Q02TptWZ0
#FOG2025
December 5, 2024 at 2:59 PM
Register now to join me at The Festival of Genomics & Biodata next month.
Registration is free for over 90% of the audience, click here to reserve your space: hubs.la/Q02TptWZ0
#FOG2025
Registration is free for over 90% of the audience, click here to reserve your space: hubs.la/Q02TptWZ0
#FOG2025
Bonus getting to catch up with old friend/colleague @pedrobeltrao.bsky.social
November 28, 2024 at 7:54 PM
Bonus getting to catch up with old friend/colleague @pedrobeltrao.bsky.social
In Zurich for a talk and participation in a course and it sure is lovely this time of year!
November 28, 2024 at 7:53 PM
In Zurich for a talk and participation in a course and it sure is lovely this time of year!
