AIEdit: Alignment-free genome assembly polisher trained on spaced seed match patterns
PLOS Computational Biology 22(5): e1014245 · May 2026
Ivana Sánchez Olivares
Bioinformatics · Machine Learning · Statistical Modelling · Reproducible Pipelines
About Me
I am a bioinformatician motivated to make a real impact on human health through data-driven insights. I build statistical models, machine learning methods, and reproducible computational pipelines to advance biomedical research and healthcare. I also enjoy science communication, interviewing epigenetics researchers for the CEEHRC podcast and helping organize its seminar series.
- Python
- R
- SQL
- PySpark
- Bash
- Machine Learning
- Deep Learning
- Single-Cell Analysis
- Genomics
- Epigenetics
- Transcriptomics
- Snakemake
- Nextflow
- Docker
- HPC
Publications
Career
-
AI Trainer in Bioinformatics, Data Science and Healthcare
Data Annotation · Freelance · Vancouver
Design bioinformatics and ML tasks with gold-standard solutions, and evaluate LLM outputs against them to generate feedback for reinforcement learning.
-
Podcast Team Member & Seminars Committee Representative
Canadian Epigenetics, Environment & Health Research Consortium (CEEHRC) · Vancouver
Interview Canadian epigenetics researchers for the CEEHRC podcast and help organize the seminar series.
Listen: S04 EP2, Interview with Dr. David Cheishvili (Spotify)
-
Research Assistant
Birol Lab · Genome Sciences Centre & BC Cancer Research Institute · Vancouver
- Simulated single-cell methylation profiles with a generalized linear mixed model in Snakemake (98% CpG recovery, MAE 0.031).
- Contributed to AIEdit, a deep-learning genome assembly polisher (human assembly QV 28.7 → 32.9).
-
M.Sc. in Bioinformatics
The University of British Columbia · Vancouver
Supervised by Dr. Inanc Birol
CONAHCYT Merit-Based Scholarship · GPA 96.2/100
Thesis: Simulation of haplotype-resolved single-cell methylation profiles
-
Data Scientist Associate
BBVA · Mexico City
Built a product engagement metric that increased product usage by 30%, and a classification model for credit eligibility.
-
Data Scientist Analyst
BBVA · Mexico City
Ensembled time-series models (e.g., ARIMA, ETS-LSTM) to forecast 90-day client income, expenses, and balances, effectively predicting 80% of clients’ amounts.
-
Intern in Mathematical Modeling and Data Validation (part-time)
Intelligence Trade, S.L. (Moondi Export) · Mexico City
Identified macroeconomic and demographic factors to rank the top 20 target countries for Spanish wine exports.
-
B.Sc. in Applied Mathematics
Universidad Nacional Autónoma de México (UNAM) · Mexico City
Gabino Barreda Medal (highest GPA in cohort) · GPA 10.00/10.00
Projects
Simulation of Haplotype-Resolved Single-Cell Methylation Profiles
Can we simulate single cells with known methylation on both haplotypes and both strands of a diploid genome, as a benchmark for alignment and methylation-calling tools, and measure how much of that signal survives bisulfite sequencing?
Methods
Tools
Normative Brain Development and Aging
Can a deep learning model chart the normal range of brain structure and connectivity across age and sex in MRI data (ABIDE), so that individuals who deviate from the norm can be identified?
Methods
Tools
Benchmarking LLM-Generated Bioinformatics Workflows
Do the bioinformatics workflows that LLMs and AI agents produce use appropriate, current, and reproducible tools?
Methods
Tools
Medulloblastoma Groups 3 and 4
How do Group 3 and Group 4 medulloblastoma tumours differ at the single-cell level, and does their immune cell composition suggest immunotherapy opportunities?
Methods
Tools
Benchmarking Single-Cell Classifiers
How accurately can machine learning and deep learning models annotate cell types in mouse brain scRNA-seq data (Tabula Muris)?
Methods
Tools
Recurrence-Free Survival in Bladder Cancer
Can a survival model improve the clinical stratification of recurrence-free survival in low-grade, stage Ta bladder cancer patients?
Methods
Tools
Cold Stress and the Pufferfish Kidney
How does low-temperature stress change the kidney transcriptome of Takifugu rubripes (Japanese pufferfish)?
Methods
Tools
Automated SNV Calling Pipeline
How can single nucleotide variant calling from raw whole-genome sequencing reads be made reproducible for population-scale genomic analysis?
Methods
Tools
Paxlovid Cost-Effectiveness Model
Is expanding Paxlovid use beyond guideline criteria cost-effective for reducing severe COVID-19 disease?
Methods
Tools