Our source code, utility scripts and links to processed data and results are all available on our lab's GitHub: lnkd.in/g7EdGYuu
Hope you enjoy reading it and reach out if you have any questions or feedback!
Big thanks to Cell Reports Methods and their editorial team for the efficient review

Our results highlight the importance of distance stratification in capturing differences in long-range loops, differences in sensitivity across different statistical models and provides overall best practices for differential HiChIP analysis.

good collaborator Katia Georgopoulos in annotation of the results, we implemented a unified framework with all different approaches to date, developed performance metrics and systematically evaluated tools/tests utilized by us and others on multiple different HiChIP datasets.

We and others have worked on this problem but realized the variability in effectiveness of different approaches across different data sets. In this work, led by Sourya Bhattacharyya (now at Empirico) and Daniela Salgado Figueroa (UCSD Bioinformatics PhD student) in my lab and with help from our

I have
NIGMS R35, impact score 12
NIHGRI R21, 4th percentile
NHGRI R01, 7th percentile (co-I)
and it seems like none will be funded. 0/3.

PO (who has been very helpful) said "Unfortunately, I do not expect this application will be selected for funding in FY25."

😭

Oh my god 😱😱 good luck Anders. These are amazing scores..We have multiple single digit %ile grants with collaborators.. one is a resubmission after multiple submissions finally making “the cut” just to have the goal post moved 😥 fingers crossed for some last minute miracle for all of us

over 1000 distinct human and mouse HiChIP samples from 152 studies plus 44 high-resolution Hi-C samples. In the paper, we demonstrate its utility for interpreting GWAS and eQTL variants through SNP-to-gene linking, identifying enriched sequence motifs and motif pairs. tinyurl.com/LoopCatalog

3 years ago we decided to compile these datasets in one place. We were fortunate to get NIH support for this, which transformed it from a local resource for our lab to a comprehensive data resource. What we named Loop Catalog is now a web-based database featuring loop calls from

The third one is work led by Joaquin Reyna (former UCSD Bioinformatics PhD student) and Kyra Fetter (former UCSD undergraduate student) with help from multiple members of our lab. Seeing the increase in the number and quality of HiChIP datasets and having developed tools for its analysis,

Knowing what controls RT, which correlates with a lot of different epigenetic/chromatin features, allows for novel ways to engineer cells with desired epigenetic programs. tinyurl.com/EMBO-erce

This highlighted master transcription factor binding sites (Oct4, Sox2, Nanog for ES cells) with contributions from transcription start sites (TSS), and a combinatorial regulation by these core elements, as the basis of RT control.

Here Jesse Turner (FSU and now at NIH) and Laura Hinojosa (UCSD Bioinformatics PhD student) with help from other co-authors dissected the previously characterized ERCEs of multiple kilobases in size down to hundreds of base pairs to map out the core parts which retain control of early replication.

The second one is a follow up of our earlier collaboration with Gilbert lab (SDBRI) where we have discovered ERCEs, regulatory elements that control early replication of DNA.

This enabled us to understand the contribution of different features to high-resolution RT both locally and globally. Our comparison with not-so-deep learning methods showed that they readily provide a pretty good baseline suggestion non-linear effects may be minimal. tinyurl.com/Soffritto

One of these is more expensive than the other, you can guess. And accordingly, there is (much) more available data from one than the other. Dante developed a deep learning framework that utilizes 2-fraction RT data with histone modifications to predict 16-fraction data with pretty good accuracy.

The first one is Soffritto, developed by Dante Bolzan (UCSD Bioinformatics PhD student). When our DNA replicates, it does not do it simultaneously across all regions. One can map this replication timing (RT) using low- (two fraction – early vs late) or high-resolution (up to 16 fractions) methods.