In vivo single-cell ribosome profiling reveals cell-type-specific translational programs during aging
Dure C, GhoshDastider U et al., Sendoel A
2nd authorBioinformatics & computational biology consultant · Zurich
Dr. Umesh GhoshDastider
I help research and pharma teams turn large multi-omics and clinical datasets into clear, actionable results. I have over 10 years of experience in oncology, machine learning and cancer genomics.
From single-cell CRISPR screens to Singapore's flagship clinical genomics portal, I turn complex, multi-modal data into translational insights for cancer, obesity and diabetes.
10+Years of experience
40+Publications
27h-index
~3300Citations
About
Bridging computation and clinical impact
I am a computational biologist and data scientist with over 10 years of international experience. My work covers machine learning, oncology and multi-omics. I am based in Zurich.
I have built computational platforms, including pioneering in vivo single-cell CRISPR screening methods that speed up functional target validation. I am skilled at turning large, multi-modal datasets such as TCGA and UK Biobank into clinical hypotheses.
I work well across disciplines. I connect dry-lab computation with wet-lab experiments, turning complex biological data into better medicines.
Translational oncologyMulti-omicsAI & deep learningCloud & HPCResearch leadership
Services
Consulting services
Bioinformatics and data science consulting for research and pharma teams: four service areas, from AI strategy to cloud pipelines.
Raw readsAlign & QCQuantifyDeconvolveTrain modelsInsight
AI & machine learning strategy
- Generative AI & LLMsFine-tune large language models and transformers for biological sequences and molecular property prediction.
- Predictive modellingBayesian inference, deep learning (PyTorch, TensorFlow) and high-dimensional statistics for biomarker discovery.
- Integrative multi-omicsCombine scRNA-seq, spatial transcriptomics, proteomics and genomics to map the tumour microenvironment.
Cancer genomics & translational oncology
- Target identificationValidate new therapeutic targets with CRISPR/Cas9 screens and functional genomics.
- Precision medicineDeconvolution algorithms and polygenic risk scores to stratify patients for precision treatment.
- Clinical data miningSearch large clinical cohorts for signatures of drug resistance and treatment response.
Data engineering & cloud architecture
- Scalable pipelinesReproducible, production-grade workflows with Nextflow, Snakemake and Docker or Singularity.
- Cloud & HPC infrastructureLarge-scale analysis on AWS, Google Cloud and high-performance computing (HPC).
- Software engineeringOptimised Python and R code, with Git and CI/CD for reliable software.
Research leadership & collaboration
- Cross-functional partnershipI bridge computational teams, experimental biologists and clinical leads.
- Project leadershipLead multi-institution collaborations (ETH, UZH, industry), including patent filings.
- Scientific communicationExplain complex findings to executives and publish in high-impact, peer-reviewed journals.
Expertise
Core capabilities
From biomarker discovery to the cloud infrastructure that runs it, six areas of hands-on expertise.
Biomarker & drug discovery
- Biomarker identificationStatistical and AI/ML methods on large datasets (TCGA, CPTAC, GWAS) to find novel therapeutic targets and predictive biomarkers for cancer, obesity and diabetes.
- Drug repurposingIn silico screening to identify existing drugs that could be repurposed for new indications.
- Translational researchWorking with clinicians and pharmaceutical partners to move computational findings toward clinical application.
NGS & bioinformatics pipelines
- Pipeline developmentRobust, scalable and reproducible NGS pipelines built with Snakemake and Nextflow.
- Single-cell analysisscRNA-seq and scATAC-seq processing with Seurat, Scanpy and Monocle.
- Genomics & transcriptomicsRNA-seq, spatial transcriptomics, ChIP-seq and ATAC-seq: alignment, quantification and differential expression.
- Variant analysisVariant calling (GATK, Mutect), annotation (VEP) and interpretation.
Machine learning & AI
- Deep learningKeras/TensorFlow and PyTorch models, including Transformers, CNNs and autoencoders for genomic sequence and image data.
- Supervised learningRandom Forests, gradient boosting (XGBoost, LightGBM) and SVMs.
- Unsupervised learningClustering and dimensionality reduction (PCA, UMAP, t-SNE) for data exploration and subtype identification.
- Statistical modellingLinear and non-linear models, Bayesian inference, time-series analysis and Cox survival models.
Python & R programming
- Python stackNumPy, SciPy, Pandas, scikit-learn and deep learning libraries (Keras, PyTorch, JAX).
- R environmentTidyverse for data manipulation and visualisation, plus the Bioconductor ecosystem for genomic analysis.
- Clean codeEfficient, modular and maintainable code that follows best practice and is version-controlled with Git.
Web apps & data visualisation
- Interactive dashboardsData exploration tools built with Dash (Plotly), Shiny (R), and JavaScript, HTML and CSS.
- Web backendLightweight backend APIs with Flask, deployed on WordPress or custom servers.
- Publication-quality plotsStatic and interactive figures with ggplot2, Matplotlib, Seaborn and Plotly.
Cloud & DevOps
- Cloud computingAWS (EC2, S3, RDS) and Google Cloud for scalable data storage and high-performance computing.
- ContainerisationReproducible analysis environments with Docker and Singularity.
- Version controlGit and GitHub/GitLab for collaborative code development.
- HPCLarge-scale bioinformatics workflows on SLURM and LSF job schedulers.
Highlights
Selected results
Nature
In vivo single-cell CRISPR screens
Co-first author on a study of distinct TNF-α programmes in clonal expansion and tumour formation (2024).
Functional genomics25,000+
Patient samples on OncoSG
I started Singapore's flagship clinical genomics portal for cancer precision medicine. It serves 100+ users.
Precision medicine500,000+
Samples in large-scale AI analysis
I applied statistical and AI methods to cohorts such as UK Biobank and TCGA at ETH Zurich, with Boehringer Ingelheim.
Population & clinical dataPatent
Quantifying molecular activity in tumours
European patent EP3867403A4 (2022): a method to quantify molecular activity in the cancer cells of a human tumour.
Intellectual property50,000+
Pan-cancer multi-omic samples
I built machine learning models in Python and R, using cancer genomes, proteomes and clinical data from TCGA, CPTAC and DepMap at the University of Zurich.
Pan-cancer analytics20,000+
Patient proteogenomic datasets
At A*STAR, I analysed multi-omic data to decode tumour heterogeneity and find molecular drivers of acquired drug resistance.
Drug resistanceAlgorithm
RNA-seq deconvolution
I engineered a high-precision method to resolve tumour microenvironment complexity. It is part of a scalable Python and R ecosystem.
MethodsMentor
Master’s and PhD candidates
I co-supervised researchers at ETH Zurich, University of Zurich, NUS and A*STAR in high-dimensional data analysis.
MentorshipTechnology stack
Tools & technologies
analysis.sh
$ nextflow run main.nf -profile aws
$ python analysis.py
import scanpy as sc
adata = sc.read_10x_mtx("filtered/")
sc.pp.normalize_total(adata); sc.pp.log1p(adata)
sc.tl.umap(adata)analysis.R
library(Seurat) seu <- CreateSeuratObject(counts = counts) seu <- NormalizeData(seu) seu <- FindVariableFeatures(seu) seu <- RunPCA(seu)
model.py
from sklearn.ensemble import RandomForestClassifier from sklearn.model_selection import cross_val_score clf = RandomForestClassifier(n_estimators=500) scores = cross_val_score(clf, X, y, cv=5) print(scores.mean())
AI & Data Science
PythonRNumPyPandasscikit-learnPyTorchTensorFlow
Bioinformatics & Multi-omics
SeuratScanpyDESeq2STARGWAS (PLINK)Cell RangerHail
Cloud, Pipelines & Data
NextflowSnakemakeDockerAWSGoogle CloudPostgreSQLGit
Apps, Visualisation & Structure
DashShinyFlaskStreamlitPlotlyggplot2AlphaFold
Methods
AI and Machine Learning
Statistical inference
Bayesian inferenceCox survivalLassoGLMsANOVAFDR correction
Machine learning
Random forestGradient boostingSVMUMAPt-SNEHierarchical clustering
Deep learning
TransformersCNNsRNNsLSTMs
Math & networks
PCASpectral decompositionNetwork analysisGraph algorithms
Publications
Featured publications
Ten peer-reviewed articles, where I am first to fourth author, including papers in Nature, Molecular Cell and Cell Reports.
27h-index
38i10-index
40+Publications
~3300Citations
In vivo single-cell CRISPR uncovers distinct TNF-α programs in clonal expansion and tumorigenesis
Renz PF*, GhoshDastider U* et al., Sendoel A
Co-firstExploring the pan-cancer landscape of posttranscriptional regulation
GhoshDastider U, Sendoel A
1st authorA pan-cancer metabolic atlas of the tumor microenvironment
Rohatgi N, GhoshDastider U, Baruah P, Kulshrestha T, Skanderup AJ
2nd authorMonitoring the 5′UTR landscape reveals isoform switches to drive translational efficiencies in cancer
Weber R, GhoshDastider U et al., Sendoel A
2nd authorPan-cancer analysis of the ligand-receptor crosstalk in the tumor microenvironment
GhoshDastider U, Skanderup AJ
1st authorFGF-2-dependent signaling activated in aged human skeletal muscle promotes intramuscular adipogenesis
Mathes S, Fahrner A, GhoshDastider U, Ruediger HA, Leunig M, Wolfrum C, Krützfeldt J
3rd authorThe diabetes gene JAZF1 is essential for the homeostatic control of ribosome biogenesis and function in metabolic stress
Kobita A, Godbersen S, Araldi E, GhoshDastider U et al., Stoffel M
4th authorActin and tropomyosin: the evolution of compositionally and functionally distinct actin filaments
Gunning PW*, GhoshDastider U*, Whitaker S, Popp D, Robinson RC
Co-firstAction of molecular switches in GPCRs – theoretical and experimental studies
Trzaskowski B, Latek D, Yuan S, GhoshDastider U, Debinski A, Filipek S
4th authorCommon questions
How we can work together
What kinds of projects do you take on?
Multi-omics data integration, cancer genomics and biomarker discovery, machine learning for biological and clinical data, and the bioinformatics pipelines that keep results reproducible.
What data have you worked with?
Single-cell and spatial transcriptomics, proteomics, genomics and large clinical cohorts, including TCGA, CPTAC, DepMap and UK Biobank. I also started the OncoSG clinical genomics portal, which holds more than 25,000 patient samples.
Do you build software, pipelines and apps?
Both. I build reproducible Nextflow and Snakemake workflows in Docker or Singularity, and run them on AWS, Google Cloud or HPC. I also build interactive apps and dashboards in Python with Dash, Flask and Streamlit, and in R with Shiny.
Do you work with industry as well as academia?
Yes. I have worked with Boehringer Ingelheim on large-scale AI analysis. I have also collaborated with ETH Zurich, the University of Zurich, the National University of Singapore and A*STAR.
Can you help with drug discovery?
Yes. I work on target identification with CRISPR/Cas9 screens, biomarker discovery, and analysis of drug resistance and treatment response. I have also worked with Boehringer Ingelheim on AI analysis.
Can you work with sensitive clinical data?
Yes. I have worked with large clinical and population cohorts, including UK Biobank and the patient data behind the OncoSG clinical genomics portal. My pipelines run in Docker or Singularity containers, so every analysis can be rerun and audited.
Contact
Start a consulting project
I am open to consulting projects, research collaborations and full-time roles. Email me a short note about your dataset, your question and your timeline.