#64 Enformer: predicting gene expression from sequence with Žiga Avsec

In this episode, Marie Sadler talks about her recent Cell Genomics paper, Multi-layered genetic approaches to identify approved drug targets. Previous studies have found that the drugs that target a gene linked to the disease are more likely to be approved. Yet there are many ways to define what it means for a gene to be linked to the disease. Perhaps the most straightforward approach is to rely on the genome-wide association studies (GWAS) data, but that data can also be integrated with quantitative trait loci (eQTL or pQTL) information to establish less obvious links between genetic variants (which often lie outside of genes) and genes. Finally, there’s exome sequencing, which, unlike GWAS, captures rare genetic variants. So in this paper, Marie and her colleagues set out to benchmark these different methods against one another. Listen to the episode to find out how these methods work, which ones work better, and how network propagation can improve the prediction accuracy. Links: Multi-layered genetic approaches to identify approved drug targets (Marie C. Sadler, Chiara Auwerx, Patrick Deelen, Zoltán Kutalik) Marie on GitHub Interview with Mariana Mamonova, the Ukrainian marine infantry combat medic who spent 6 months in russian captivity while pregnant Thank you to Jake Yeung, Michael Weinstein, and other Patreon members for supporting this episode.

Today on the podcast we have Tomasz Kociumaka and Dominik Kempa, the authors of the preprint Collapsing the Hierarchy of Compressed Data Structures: Suffix Arrays in Optimal Compressed Space. The suffix array is one of the foundational data structures in bioinformatics, serving as an index that allows fast substring searches in a large text. However, in its raw form, the suffix array occupies the space proportional to (and several times larger than) the original text. In their paper, Tomasz and Dominik construct a new index, δ-SA, which on the one hand can be used in the same way (answer the same queries) as the suffix array and the inverse suffix array, and on the other hand, occupies the space roughly proportional to the gzip’ed text (or, more precisely, to the measure δ that they define — hence the name). Moreover, they mathematically prove that this index is optimal, in the sense that any index that supports these queries — or even much weaker queries, such as simply accessing the i-th character of the text — cannot be significantly smaller (as a function of δ) than δ-SA. Links: Collapsing the Hierarchy of Compressed Data Structures: Suffix Arrays in Optimal Compressed Space (Dominik Kempa, Tomasz Kociumaka) Thank you to Jake Yeung and other Patreon members for supporting this episode.

In this episode, David Dylus talks about Read2Tree, a tool that builds alignment matrices and phylogenetic trees from raw sequencing reads. By leveraging the database of orthologous genes called OMA, Read2Tree bypasses traditional, time-consuming steps such as genome assembly, annotation and all-versus-all sequence comparisons. Links: Inference of phylogenetic trees directly from raw sequencing reads using Read2Tree (David Dylus, Adrian Altenhoff, Sina Majidian, Fritz J. Sedlazeck, Christophe Dessimoz) Background story Read2Tree on GitHub OMA browser The Guardian’s podcast about Victoria Amelina and Volodymyr Vakulenko

This is the third and final episode in the AlphaFold series, originally recorded on February 23, 2022, with Amelie Stein, now an associate professor at the University of Copenhagen. In the episode, Amelie explains what 𝛥𝛥G is, how it informs us whether a particular protein mutation affects its stability, and how AlphaFold 2 helps in this analysis. A note from Amelie: Something that has happened in the meantime is the publication of methods that predict 𝛥𝛥G with ML methods, so much faster than Rosetta. One of them, RaSP, is from our group, while ddMut is from another subset of authors of the AF2 community assessment paper. Other links: A structural biology community assessment of AlphaFold2 applications (Mehmet Akdel, Douglas E. V. Pires, Eduard Porta Pardo, Jürgen Jänes, Arthur O. Zalevsky, Bálint Mészáros, Patrick Bryant, Lydia L. Good, Roman A. Laskowski, Gabriele Pozzati, Aditi Shenoy, Wensi Zhu, Petras Kundrotas, Victoria Ruiz Serra, Carlos H. M. Rodrigues, Alistair S. Dunham, David Burke, Neera Borkakoti, Sameer Velankar, Adam Frost, Jérôme Basquin, Kresten Lindorff-Larsen, Alex Bateman, Andrey V. Kajava, Alfonso Valencia, Sergey Ovchinnikov, Janani Durairaj, David B. Ascher, Janet M. Thornton, Norman E. Davey, Amelie Stein, Arne Elofsson, Tristan I. Croll & Pedro Beltrao) A crime in the making: Russia’s atrocities — the podcast episode about the Olenivka prison massacre

This is the second episode in the AlphaFold series, originally recorded on February 14, 2022, with Janani Durairaj, a postdoctoral researcher at the University of Basel. Janani talks about how she used shape-mers and topic modelling to discover classes of proteins assembled by AlphaFold 2 that were absent from the Protein Data Bank (PDB). The bioinformatics discussion starts at 03:35. Links: A structural biology community assessment of AlphaFold2 applications (Mehmet Akdel, Douglas E. V. Pires, Eduard Porta Pardo, Jürgen Jänes, Arthur O. Zalevsky, Bálint Mészáros, Patrick Bryant, Lydia L. Good, Roman A. Laskowski, Gabriele Pozzati, Aditi Shenoy, Wensi Zhu, Petras Kundrotas, Victoria Ruiz Serra, Carlos H. M. Rodrigues, Alistair S. Dunham, David Burke, Neera Borkakoti, Sameer Velankar, Adam Frost, Jérôme Basquin, Kresten Lindorff-Larsen, Alex Bateman, Andrey V. Kajava, Alfonso Valencia, Sergey Ovchinnikov, Janani Durairaj, David B. Ascher, Janet M. Thornton, Norman E. Davey, Amelie Stein, Arne Elofsson, Tristan I. Croll & Pedro Beltrao) The Protein Universe Atlas What is hidden in the darkness? Deep-learning assisted large-scale protein family curation uncovers novel protein families and folds (Janani Durairaj, Andrew M. Waterhouse, Toomas Mets, Tetiana Brodiazhenko, Minhal Abdullah, Gabriel Studer, Mehmet Akdel, Antonina Andreeva, Alex Bateman, Tanel Tenson, Vasili Hauryliuk, Torsten Schwede, Joana Pereira) Geometricus: Protein Structures as Shape-mers derived from Moment Invariants on GitHub The group page The Folded Weekly newsletter A New York Times article about the Kramatorsk missile strike. The Instagram video, part of which you can hear at the beginning of the episode, appears to have been deleted.

In this episode, originally recorded on February 9, 2022, Roman talks to Pedro Beltrao about AlphaFold, the software developed by DeepMind that predicts a protein’s 3D structure from its amino acid sequence. Pedro is an associate professor at ETH Zurich and the coordinator of the structural biology community assessment of AlphaFold2 applications project, which involved over 30 scientists from different institutions. Pedro talks about the origins of the project, its main findings, the importance of the confidence metric that AlphaFold assigns to its predictions, and Pedro’s own area of interest — predicting pockets in proteins and protein-protein interactions. Links: A structural biology community assessment of AlphaFold2 applications (Mehmet Akdel, Douglas E. V. Pires, Eduard Porta Pardo, Jürgen Jänes, Arthur O. Zalevsky, Bálint Mészáros, Patrick Bryant, Lydia L. Good, Roman A. Laskowski, Gabriele Pozzati, Aditi Shenoy, Wensi Zhu, Petras Kundrotas, Victoria Ruiz Serra, Carlos H. M. Rodrigues, Alistair S. Dunham, David Burke, Neera Borkakoti, Sameer Velankar, Adam Frost, Jérôme Basquin, Kresten Lindorff-Larsen, Alex Bateman, Andrey V. Kajava, Alfonso Valencia, Sergey Ovchinnikov, Janani Durairaj, David B. Ascher, Janet M. Thornton, Norman E. Davey, Amelie Stein, Arne Elofsson, Tristan I. Croll & Pedro Beltrao) Pedro’s group at ETH Zurich

In this episode, Jacob Schreiber interviews Žiga Avsec about a recently released model, Enformer. Their discussion begins with life differences between academia and industry, specifically about how research is conducted in the two settings. Then, they discuss the Enformer model, how it builds on previous work, and the potential that models like it have for genomics research in the future. Finally, they have a high-level discussion on the state of modern deep learning libraries and which ones they use in their day-to-day developing. Links: Effective gene expression prediction from sequence by integrating long-range interactions (Žiga Avsec, Vikram Agarwal, Daniel Visentin, Joseph R. Ledsam, Agnieszka Grabska-Barwinska, Kyle R. Taylor, Yannis Assael, John Jumper, Pushmeet Kohli & David R. Kelley ) DeepMind Blog Post (Žiga Avsec)

The Bioinformatics Contest is back this year, and we are back to discuss it! This year’s contest winners Maksym Kovalchuk (1st prize) and Matt Holt (2nd prize) talk about how they approach participating in the contest and what strategies have earned them the top scores. Timestamps and links for the individual problems: 00:10:36 Genotype Imputation 00:21:26 Causative Mutation 00:30:27 Superspreaders 00:37:22 Minor Haplotype 00:46:37 Isoform Matching Links: Matt’s solutions Max’s solutions

In this episode, Apostolos Chalkis presents sampling steady states of metabolic networks as an alternative to the widely used flux balance analysis (FBA). We also discuss dingo, a Python package written by Apostolos that employs geometric random walks to sample steady states. You can see dingo in action here. Links: Dingo on GitHub Searching for COVID-19 treatments using metabolic networks Tweag open source fellowships This episode was originally published on the Compositional podcast.

In this episode, Jacob Schreiber interviews Da-Inn Erika Lee about data and computational methods for making sense of 3D genome structure. They begin their discussion by talking about 3D genome structure at a high level and the challenges in working with such data. Then, they discuss a method recently developed by Erika, named GRiNCH, that mines this data to identify spans of the genome that cluster together in 3D space and potentially help control gene regulation. Links: GRiNCH: simultaneous smoothing and detection of topological units of genome organization from sparse chromatin contact count matrices with matrix factorization (Da-Inn Lee and Sushmita Roy) GRiNCH Project Page In silico prediction of high-resolution Hi-C interaction matrices (Shilu Zhang, Deborah Chasman, Sara Knaack, and Sushmita Roy)