Journal Club
Reading the literature, out loud.
Deep, critical reads of the papers shaping non-coding cancer genomics, regulatory deep learning and 3D genome biology, each one tied back to what I'm building in my PhD, with animated explainers for the hard bits.
Mini Journal Club · Non-coding variants
From enhancer networks to base-resolution variant scoring
A mini journal club on three papers about non-coding regulatory variants: Wang & Hon's single-cell CRISPRi enhancer-to-gene network in breast cancer, Gerasimavicius' AlphaFold 3 structural scoring of transcription-factor binding, and Marderstein & Montgomery's FLARE, which prioritises variants across 132 cell types using chromatin deep learning plus evolutionary constraint.
Read the analysisMini Journal Club · 3D genome
From correlation to contact: how enhancers really find their genes
A mini journal club on three papers behind my Curie contact work: the Activity-by-Contact model, the 3D genome landscape of primary human cancers, and SCEG-HiC, the principle, the tumour data, and the single-cell engine for contact-supported enhancer-gene links.
Mini Journal Club · Breast cancer genomics
From healthy breast to lncRNA drivers: two papers, two ends of my thesis
A mini journal club on the Human Breast Cell Atlas and a pan-cancer screen for driver mutations in long non-coding RNAs, how a single-cell reference and 12,631 tumour genomes bracket the exact problem my PhD sits in.
Mini Journal Club · Regulatory deep learning
Two ways to learn regulatory grammar: personal genomes vs MPRAs
A mini journal club on Variformer and PARM, fine-tuning Enformer on personal genomes versus training a lightweight CNN on millions of MPRA fragments, and what each says about whether deep learning has actually learned the grammar of gene regulation.
Journal Club · Cancer evolution
Selection isn't causation: using patient age to find real cancer drivers
A journal club on Cheek et al., why a mutation under positive selection in normal tissue isn't necessarily a cancer driver, and how patient age can separate the two without any normal-tissue data.
Journal Club · Epigenetic memory
When the transcriptome forgets but the epigenome remembers
A journal club on Nagaraja et al., how chronic colitis leaves a durable, clonally inherited epigenetic memory of AP-1 activity in colonic stem cells that persists for months and primes the tissue for tumour growth.
Journal Club · Regulatory deep learning
AlphaGenome: one model to read the regulatory genome
A critical read of DeepMind's AlphaGenome, a unified sequence-to-function model that predicts 11 regulatory modalities across a 1 Mb window, and what it means for non-coding variant interpretation in cancer.
Journal Club · Review
A decade in the non-coding genome (2015–2025)
A journal-club tour of ten years of non-coding driver research in cancer, from the first 50 bp hotspot scans and the TERT exception, through PCAWG's 'paucity' verdict and GC-bias blind spots, to genome-wide and single-cell methods, and why the field is now pivoting from statistics to deep learning.
Journal Club · Non-coding drivers
Ultraconserved elements as a new class of non-coding cancer drivers
A journal club on Bayraktar et al., the mutational landscape of noncoding ultraconserved elements in human cancers, where sequences 100% identical across mammals turn out to act as mutated enhancers, silencers and miRNA-processing sites in tumours.