iiinova — where your digital self keeps evolving

iiinova is part of the Atmosphere — an open, public network where you own your identity.

Thomas Sandmann

@thomas-sandmann.genomic.social.ap.brid.gy

Computational biologist, biotech, neurodegeneration, genomics, learning, R, python [bridged from genomic.social/@thomas_sandmann on the fediverse by fed.brid.gy ]

Posts Replies Media

Thomas Sandmann @thomas-sandmann.genomic.social.ap.brid.gy · 14/10/2026

I am learning about recent methods for single-cell RNA-seq analysis. The open access review "**Representation learning of single-cell RNA-seq data**" by @const_ae, @janlause and colleagues is great! Very helpful to see so many methods positioned within clear […] [Original post on genomic.social]

Figure 2 by Constantin Ahlmann-Eltze, Florian Barkmann, Jan Lause, Valentina Boeva and Dmitry Kobak3: Conceptual illustrations of the four families of representation learning methods. (KL divergence) Kullback–Leibler divergence, (NB) negative binomial distribution, (ZINB) zero-inflated negative binomial distribution, (CLS) classification token.

001 1

Thomas Sandmann @thomas-sandmann.genomic.social.ap.brid.gy · 13/07/2025

@erc_bk My analysis report is generated from a Quarto markdown document, which contains cells with R, python or SQL code (raw.githubusercontent.com/tomsing1/…). The `db_path` and `work_dir` variables are […] [Original post on genomic.social]

A screenshot from the rmarkdown documentation, illustrating how R variables can be passed into SQL queries by prefacing the R variable name with a question mark in the SQL query.

010 1

Thomas Sandmann @thomas-sandmann.genomic.social.ap.brid.gy · 08/03/2025

Great crowd of angry Scientists and Science supporters at #standupforscience in San Francisco this afternoon. Speakers included scientists (among them Nobel laureate Carolyn Bertozzi), cancer survivors, students and California State Senator Scott Wiener. It […] [Original post on genomic.social]

099 9

Thomas Sandmann @thomas-sandmann.genomic.social.ap.brid.gy · 27/11/2024

Just read "ChIP-DIP maps binding of hundreds of proteins to DNA simultaneously and identifies diverse gene regulatory elements" by Perez et al (published in Nature Genetics, preprint is available here […] [Original post on genomic.social]

Figure 1 from the preprint; ChIP-DIP: A highly multiplexed method for mapping proteins to genomic DNA.
(A) Schematic of the ChIP-DIP method. Beads are coupled with an antibody and associated oligonucleotide (antibody-ID). Sets of beads are then mixed (antibody-bead pool) and used to perform ChIP. Multiple rounds of split-and-pool barcoding are performed to identify molecules bound by the same Protein G bead. DNA is sequenced and genomic DNA and antibody oligos containing the same split-and-pool barcode are grouped into a cluster, which are used to assign genomic DNA regions to their linked antibodies. All DNA reads corresponding to the same antibody are used to generate protein-localization maps. (B) Protein localization maps over a specific human genomic region (hg38, chr12:53,649,999-54,650,000) for four protein targets - CTCF, H3K4me3, RNAP II and H3K27me3. Left panel: Protein localization generated by ChIP-DIP in K562. Right panel: ChIP-Seq data generated by ENCODE within K562 for these same 4 proteins are shown for the same region. (C) Comparison of ChIP-DIP and ChIP-Seq maps over specific regions corresponding to zoom-ins of the larger region shown in (B). (D) Genome-wide comparison (density plots of signal correlation) between the localization of each individual protein measured by ChIP-DIP (x-axis) or ChIP-Seq (y-axis). Points are measured genome-wide across 10kb windows (CTCF, H3K27me3) or all promoter intervals (H3K4me3, RNAP II).

011 1