01 — About

Basanagowda Patil.

Bioinformatics · Computational Biology

Bioinformatics, measured against data.

Bioinformatics, as I practise it, means taking a biological question, pulling the relevant sequences or structures from a public database, and answering it with statistics rather than intuition. My main project asks whether intrinsic disorder in human protein kinases tracks protein length or charge–hydropathy. Length turns out not to matter; the classical charge–hydropathy threshold turns out to under-fire badly on multidomain enzymes, and I traced why.

I am in my third year at Marwadi University. I work in Python, I read structures from PDB and sequences from UniProt, and I have published a random-forest model of glioblastoma survival built on TCGA clinical and genomic features. I am now building toward structure-based work — docking and binding-site analysis. I would rather publish a small honest negative result than a large confident one.

Education

B.Tech Bioinformatics
Marwadi University, Rajkot

Year

Third year
Expected graduation 2028

Publication

IEEE ASSIC 2026
GBM survival · Random Forest + SHAP

Experience

Software engineering intern, Bluestock
Python in practice

Leadership

Secretary, IEEE EMBS Student Chapter
Genomics Club — industry expert talks

Award

Code Carnival 3.0
Third place

02 — Work

Research and projects, with the numbers.

Survival prediction in glioblastoma

IEEE ASSIC 2026 · Random Forest · SHAP

Glioblastoma is the most aggressive primary brain tumour in adults, and survival is hard to predict from conventional methods alone. I built a framework that predicts overall survival in months from integrated clinical and molecular features of the TCGA dataset.

Preprocessing — missing-value imputation, feature scaling and one-hot encoding — then a Random Forest trained on survival in months.
Low mean squared error despite the biological variability of GBM; the models reached 87% and 84% accuracy respectively.
Feature importance and SHAP identified Disease Free (Months), Diagnosis Age and Fraction Genome as the influential variables, and confirmed the predictions rest on clinically relevant features rather than artefacts.

Published in the 2026 International Conference on Advancements in Smart, Secure and Intelligent Computing (ASSIC), Bhubaneswar, India, 8–10 August 2026. DOI 10.1109/ASSIC70037.2026.11717077.

IEEE Xplore DOI

kinase-disorder-scan

Python · Jupyter · UniProt REST

Does intrinsic disorder in human protein kinases track protein length or charge–hydropathy? Pulled all reviewed human kinases from UniProtKB/Swiss-Prot (500 of 625 retrieved) and analysed the 320 carrying curated disordered-region annotations from DisProt/MobiDB.

Disorder content: median 12.1% of residues (IQR 6.6–21.3%, range 1.1–64.3%).
Length does not predict disorder — Pearson r = −0.003, p = 0.95. Charge–hydropathy tracks it continuously: r = −0.552, p = 6.7 × 10⁻²⁷.
The classical binary threshold fails as a classifier — accuracy 0.491, sensitivity 0.138, specificity 0.992, MCC 0.232. The ordered catalytic domain dominates the whole-protein average and pushes the score above the boundary.

Limitations written into the repo: the method was calibrated on small globular proteins and isolated disordered proteins; UniProt annotations change over time; one misclassification was left un-tuned rather than fixed after the fact.

Repository

hackerrank-orchestrate

Python

An AI agent that audits whether a user's personal data was actually deleted. It decides whether deletion was requested, then returns COMPLETE, INCOMPLETE or NOT_APPLICABLE with remaining-data and recreation-risk findings and ordered remediation steps. Includes a classifier module, an evaluation harness and written reports. HackerRank Orchestrate hackathon, October 2026.

Repository

Geleya

React · TypeScript

An AI wellness companion with an AI Stress Score and an Academic Load Tracker. Built and deployed. Hackathon project.

Live site Repository

drivedeck

TypeScript

Placement-prep web app with a kanban application board and an auto-built public portfolio, backed by a PRD, an architecture document and a design document. JioCoder Hackathon 2026.

Repository