Sasha Chernenkoff

Fine-tuning genomic foundation models for protein abundance prediction University of Calgary MSc Thesis Project
Protein abundance is shaped by a complex interaction between transcription, post-transcriptional control, translation efficiency, and post-translational control. These factors are difficult to learn from limited labeled data, since mRNA levels alone are a weak proxy for protein output. This project fine-tunes a pretrained genomic foundation model trained on large-scale regulatory genomics data as a foundation for protein abundance prediction. Rather than training from scratch, it leverages learned representations of DNA regulatory context to test whether pretrained sequence-to-function models can generalize beyond their original training tasks to a distinct molecular phenotype.

Diploid, patient-specific gene expression modelling University of Calgary Research Project
Most genomic sequence-to-function models operate on haploid reference sequences. This makes them poor predictors of real patient regulatory variation. They fail to capture cis-epistasis: the non-additive compounding effect of multiple regulatory variants interacting on the same haplotype. This is a limitation for models intended for clinical use. This projects uses diploid, patient-specific genome sequences to predict personalized gene expression coverage. It finetunes the trunk of a genomic foundation model and trains a new prediction head. The project investigates how much of the model's improvement on personalized data reflects genuine genotype-driven signal rather than simply learning a new data distribution.

VAE-based heritability estimation from genotype data University of Calgary Research Project
Standard heritability estimation methods, such as GCTA-GREML, rely on genetic relationship matrices (GRMs) built directly from genotype data or from linear dimensionality-reduction techniques like PCA, which may not fully capture nonlinear structure in high-dimensional genomic data. This project explores variational autoencoders (VAEs) as an alternative, nonlinear approach to genotype dimensionality reduction, learning latent representations of genotype data across multiple disease and cohort datasets. GRMs constructed from these VAE-derived latent representations are used to estimate SNP-based heritability via a GREML framework, with the goal of comparing this deep-learning-based approach against conventional PCA-based heritability estimation.

MPS I & II PerkinElmer Genomics
At PerkinElmer Genomics, I developed semi-quantitative assays to detect the presence of analytes (found to only be present in individuals with MPS I & II) from dried blood spots using LC-MS/MS, which can be run simultaneously as a panel. These markers are virtually undetectable within the normal population, yielding a larger minimum differential factor and higher confidence in results compared to traditional tests for MPS. In addition, the tests also allowed for differentiation of pseudodeficient allele variants from pathogenic variants, reducing the number of false positives commonly seen in MPS I testing. We then began to investigate trends between variant types and MPS marker concentrations, specifically for MPS I, as we believe it could prove clinically useful in understanding variants of unknown significance, which currently yield ambiguity in predicting clinical status.

Pompe CRIM-status determination PerkinElmer Genomics
At PerkinElmer Genomics, I developed a qualitative assay to detect the presence of endogenous acid α-glucosidase (GAA) within human peripheral blood mononuclear cells (PBMCs) isolated from whole blood using Western blotting. The assay allows for rapid CRIM-status determination in patients with Pompe disease in order to more quickly initiate immune tolerance induction therapy.

KDM6B 'bump-and-hole' research University of Pittsburgh Undergraduate Research
As an undergraduate at the University of Pittsburgh, I utilized the 'bump-and-hole' method to develop variants of KDM6B with 'hole-modified' active sites which were sensitive to selective inhibition by α-ketoglutarate engineered with complementary steric 'bumps' to help elucidate KDM6B's specific epigenetic function.