Make runs easier to trust
Ran RNA-seq and TWAS across ~1,100 human brain cortex samples on SLURM, resolved Nextflow memory/input/batch issues, and built Python checks for run status.
Bioinformatics Engineer / Analyst
Omics pipelines that hold up after the first run.
I’m Pranava. I turn messy sequencing projects into runnable pipelines, quality checks, and cautious biological readouts that collaborators can review.
Looking for roles where I can help teams move from raw sequencing data to checked outputs, clear reports, and well-scoped conclusions.
Proof points
~1,100 samples
51 mapped rules
63,689 cells
AUC 0.959
Evidence in:
RNA-seq/TWAS on SLURM
OpenVax Nextflow port
Python reporting
Docker mini-runtime validation
Nextflow failure triage
Single-cell annotation
Visium review
ClinVar benchmarking
What I bring
I sit between the run log and the biological question.
The work has two sides: did the computation finish cleanly, and does the result support the story being told?
Keep interpretation grounded
Across RNA-seq, spatial, single-cell, and ctDNA work, I separate observed signal from what the dataset cannot prove.
Package the handoff
I write reports, dashboards, notebooks, and notes that leave the next person with the method, output, caveats, and next checks.
Meet
Pranava
My lane is the handoff between computation and biology: get the data moving, check the output, and explain the result plainly.
At UT Dallas, I worked on RNA-seq and TWAS runs on SLURM, debugged Nextflow memory/input/batch failures, built Python status checks, and shared findings in reports and project meetings.
I am also interested in spatial transcriptomics methods development and PhD study, especially work that makes spatial results easier to validate and explain.
Earlier, I was closer to the bench: RNA-seq library prep, PCR, R/Bash scripts, antimicrobial assays, and SOP-driven microbiology. That background is why I care about both the system and the sample behind it.
Experience
The work history behind the projects.
Condensed resume evidence: HPC execution, R/Python scripting, quality reporting, and biology-facing documentation.
Bioinformatics Analyst, University of Texas at Dallas
Ran RNA-seq and TWAS across ~1,100 human brain cortex samples on Linux SLURM HPC. Debugged three recurring Nextflow failure modes: memory limits, malformed inputs, and partial batches. Built Python checks with pandas and matplotlib, then communicated findings through reports and project meetings.
- RNA-seq / TWAS
- SLURM HPC
- Nextflow debugging
- Python status checks
Bioinformatics Research Associate, Bharati Vidyapeeth University
Prepared bulk RNA-seq libraries and PCR-amplified genomic targets, then analyzed expression data in R. Built reusable R and Bash scripts across four analysis steps: quality control, normalization, differential expression, and GSEA, with annotated R Markdown reporting.
- Bulk RNA-seq
- R / Bash
- DE analysis
- R Markdown
M.S. Bioinformatics & Computational Biology
University of Texas at Dallas, May 2025. Earlier training includes a B.S. in Microbiology, Biochemistry & Chemistry from Sri Venkateshwara University, plus microbiology experience in antimicrobial susceptibility testing and SOP documentation.
- UT Dallas
- Microbiology foundation
- CLSI-style assays
- Reproducible records
Review guide
How to judge the work.
Can it be rerun?
Look for command-line entry points, parser tests, summaries, model cards, and notes. Those details matter more than a polished figure alone.
Are the claims bounded?
The spatial, single-cell, and ctDNA projects include biological context, but the conclusions stay proportional to the data and validation available.
Are baselines visible?
TinyVariant and run-risk show the habit I want to keep: start simple, check leakage, document limits, and explain the result even when the simpler method wins.
Selected work
Start with these projects.
They show my preferred pattern: make the process observable, produce a concrete artifact, and document the boundary.
Pipeline reliability / Python CLI
nf-core/rnaseq Run-Risk Predictor
Built a Python CLI/package for nf-core/rnaseq triage. It reads trace/log telemetry, labels failed or resource-risky runs, trains interpretable logistic models, and writes JSON/Markdown reports.
- Proof
- make-runs, parse, train, predict, mega-run commands
- Boundary
- Proof-of-concept scope on 51 small test-profile runs
- Evidence
- Telemetry parsing, null handling, tests, model card, and reports
With more real run history, I would test whether the risk labels stay useful outside small test-profile runs.
Open repository
Nextflow port / workflow engineering
OpenVax Nextflow Port
Ported the public OpenVax neoantigen vaccine pipeline shape from Snakemake into a Nextflow DSL2 implementation, keeping the first pass close to the original command logic before modernization.
- Proof
- Connected faithful_port workflow, startup validation, run metadata, reports, and manifests
- Runtime
- Two-patient Docker mini run with 106 succeeded Nextflow tasks
- Boundary
- Rule-level parity and local Docker validation, not clinical or cloud runtime validation
The useful part was keeping the biological logic conservative while making execution, resources, and outputs easier to review.
Open repository
Spatial transcriptomics / Nextflow / R
Spatial Insights Cohort
Designed a seven-step Nextflow/R pipeline for public DLPFC Visium data: spot filtering, normalization, domain detection, SVG ranking, neighborhood summaries, and Shiny/Plotly review.
- Scale
- 2 Visium samples, 7,783 spots
- Boundary
- Pipeline and dashboard engineering with descriptive interpretation
- Artifacts
- Validation report, step summaries, per-sample outputs, and spatial quality artifacts
Next, I would add a true SpaceRanger ingest path only if raw 10x inputs are available.
Open repository
ML evaluation / ClinVar
TinyVariant Benchmark
Built a reproducible ClinVar missense benchmark around a Tiny Recursive Model fork, then documented the important result: logistic regression remained stronger than the neural approach.
- Result
- Logistic regression AUC 0.959 vs best TRM AUC 0.951
- Signal
- Evaluation discipline over model hype
- Checks
- Preprocessing, leakage tests, ablations, sweep summary, and interpretation
What surprised me was that the simpler baseline stayed stronger, which made the negative result more useful than a cleaner model story.
Open repository
Single-cell analysis
CRC TME scRNA-seq
Processed public colorectal cancer single-cell data through quality filtering, integration, clustering, marker-informed annotation, and tumor-normal composition visualization.
With more time, I would make the annotation step less manual and keep intermediate tables beside the figures.
Open repository
Supporting work
Additional Projects
RNA-seq pipeline
Yeast oxidative stress
Nextflow, nf-core, Docker, DESeq2
Expression profiling
Pan-cancer RNA-seq
DESeq2, TCGA, survival-style summaries
Medical image ML
Gleason classifier
ResNet-50, Grad-CAM, prostate histology
Sequence simulation
SeqMorph
Mutation injection, CLI tooling
Sequence encoding
CORE-seq
Nucleotide storage, ML preprocessing
Genomics data ops
SeqOps
Scanner, index-readiness checks, dry-run recommendations
Variant review
ctDNA KRAS workflow
Depth, alt counts, VAF summaries, conservative readout
Docking study
Nociception study
Gut microbial metabolites and receptor-target exploration
Stack
The stack behind the work.
01
Pipeline systems
Nextflow, nf-core, Snakemake, CellRanger, SpaceRanger, Docker, Singularity, Conda, Git, GitHub Actions, pytest, and documentation for repeatable runs.
02
Sequencing data
RNA-seq, TWAS, scRNA-seq, snRNA-seq, ATAC-seq, 10x Visium, FASTQ/BAM/VCF/H5AD, 10x outputs, GEO, SRA, and TCGA-style public datasets.
03
Analysis libraries
Scanpy, Seurat, scVI, SpatialExperiment, BayesSpace, Moran's I, DESeq2, edgeR, limma, Harmony, and CellChat.
04
Evaluation and reporting
scikit-learn, PyTorch, logistic baselines, ROC AUC, leakage checks, R Shiny, Plotly, dashboards, and reports.
Contact