Laura Gunsalus

Excited about regulatory genomics and deep learning. ML Scientist at Genentech. Previously grad student at @ucsf in the Pollard and Keiser labs.

[Laura Gunsalus @lauragunsalus]

@MartinLozaLopez thank you, that's very kind!

[Laura Gunsalus @lauragunsalus]

We'd love to chat at our poster (#118) at #MLCB2023!

[Laura Gunsalus @lauragunsalus]

Polygraph provides a toolkit to analyze sequence and k-mer composition, motif abundance, and predicted activity with pre-trained models. Polygraph also uses a human DNA language model to quantify the log-likelihood of synthetic sequences and score their "humanness".

[Laura Gunsalus @lauragunsalus]

We apply Polygraph to two datasets of synthetic yeast promoters and human enhancers. Polygraph is the first instrument for assessing synthetic regulatory sequences, enabling progress in therapeutic applications and improving our understanding of gene regulatory mechanisms.

[Laura Gunsalus @lauragunsalus]

We can now computationally engineer regulatory elements to drive cell-type specific expression, but it’s still difficult to know which mechanisms they exploit, if they converge on the same grammar as endogenous enhancers and promoters, and which sequences to test in the lab.

[Laura Gunsalus @lauragunsalus]

We’re excited to share Polygraph, a Python framework for evaluating native and synthetic DNA regulatory elements! @lal_avantika @gokcen https://t.co/mJixAWc2Mb

[Laura Gunsalus @lauragunsalus]

We find that NMF is a super simple and effective way to link the chromatin structure in individual cells to nascent transcription⚡️

[antisense. @razoralign]

ChromaFactor: deconvolution of single-molecule chromatin organization with non-negative matrix factorization https://t.co/DsLio6bONk

[Laura Gunsalus @lauragunsalus]

More info here! https://t.co/Ts8Eahjwf4

[Laura Gunsalus @lauragunsalus]

Excited to share our work developing experimental strategies for deep learning models to decipher how DNA sequence informs 3D genome folding! Our study points to understudied links between repetitive elements and chromatin organization. Link: https://t.co/CtcU7EbjMS (1/12)

[Laura Gunsalus @lauragunsalus]

Hooray, it's out! We use a deep learning model to screen the genome and find non-coding RNAs, Alus, SVAs, other transposable elements as sequence determinants of 3D genome folding.

[Cell Genomics @CellGenomics]

In silico discovery of repetitive elements as key sequence determinants of 3D genome folding https://t.co/9xgwMrSd1W

[Elphege Nora Lab at UCSF @ElphegeNoraLab]

New insight on how to compare two Hi-C maps quantitatively

TLDR - not trivial at all!
Interpretations can greatly depend on method choice

From @lauragunsalus + @EvonneMcArthur

[Laura Gunsalus @lauragunsalus]

@ElphegeNoraLab @EvonneMcArthur Thanks Elphege!! This was such a fun collaborative project that sprung out of a problem we all kept running into.

[Laura Gunsalus @lauragunsalus]

@XinyangBing I think it's still hard to draw conclusions about especially recent, near-identical elements though!

[Laura Gunsalus @lauragunsalus]

Excited to share our work developing experimental strategies for deep learning models to decipher how DNA sequence informs 3D genome folding! Our study points to understudied links between repetitive elements and chromatin organization. Link: https://t.co/CtcU7EbjMS (1/12)

[Laura Gunsalus @lauragunsalus]

@XinyangBing To your second point- yes this is a legitimate concern. We looked at sequence mappability and didn't find a correlation with our measure of genome sensitivity (check out S4 and the supplemental note for more info).

[Laura Gunsalus @lauragunsalus]

Overall, we highlight a diverse vocabulary of repetitive elements that collaborate with CTCF to inform genome folding. Thanks to Katie Pollard, @keiser_lab, @gfudenberg + @drklly, whose work we build on, and the many others who gave feedback along the way. @4DNucleome (12/12)

[Laura Gunsalus @lauragunsalus]

…and an in silico dissection of which sequence is necessary and sufficient for a CTCF-independent Charlie7 element to make a boundary. (11/12)