Cheng Soon Ong

Data-Efficient Exploration of Enzyme Function Using Family-Specific Machine Learning

The paper combines experimental screening with sequence-based deep learning to efficiently discover high-performing enzyme variants.

Generative Machine Learning and Microfluidics uHTS: An Efficient Partnership for Enzyme Engineering

This paper combines microfluidic high-throughput screening with a generative model to engineer an enzyme with improved specificity, achieving higher product enrichment than traditional screening methods.

Machine learning guided batched design of a bacterial Ribosome Binding Site

This study uses machine learning to design bacterial ribosome binding sites (RBSs), achieving high translation initiation rates through iterative experimental testing.

A novel haplotype-based eQTL approach identifies genetic associations not detected through conventional SNP-based methods

The paper introduces eQTLHap, a haplotype-based eQTL method that detects genetic associations missed by SNP-based approaches, finding 22 replicated associations in real datasets.

Machine learning techniques for classifying the mutagenic origins of point mutations

This paper uses machine learning to classify the mutagenic origins of point mutations by analyzing sequence context, showing that neighboring bases can distinguish between ENU-induced and spontaneous mutations.

Disentangled behavioral representations

The paper develops a framework to represent individual differences in decision-making using low-dimensional, interpretable latent spaces derived from recurrent neural networks.

kWIP: The k-mer Weighted Inner Product, a de novo Estimator of Genetic Similarity

kWIP estimates genetic similarity de novo using k-mer counts, enabling unbiased detection of sample mix-ups and population structure without reference genomes.