Yari Ciani

In light of our results, we support the implementation of the T2T-CHM13 reference for the improvement of sequencing data analyses in the clinical genomic setting. DemichelisLab (@FrancescaBZNY) @CIBIO_UniTrento @UniTrento

· Reference genome assembly choice impacts the detection of clinically relevant variants , including pathogenic variants in BRCA1 gene

Does changing the reference genome from hg38 to T2T-CHM13 affect mapping over clinically relevant variants? We show that the T2T-CHM13 reference genome reduces bias in human sequence analysis with implications for variant discovery and clinical genomics.biorxiv.org/content/10.648…

Favorite study! Germline-somatic tandems dictate cancer subtypes, provide clinical insights, and fuel molecularly informed studies. Kudos: Blerta Stringa, @Paola_Gasperini, Alessandro Alaimo, and @LunardiLabCIBIO (co-last a!) @NimaSharifiMD doi.org/10.1158/2159-8… @UniTrento

🚨 Excited to announce our new paper, “SpikeFlow: Automated and Flexible Analysis of ChIP-Seq Data with Spike-in Control”, now online in NAR Genomics and Bioinformatics! 🚨 🔗 academic.oup.com/nargab/article… @LabChiacchiera @aromanel

🔥Post-doc position open🔥 at the Demichelis laboratory @UniTrento ⏳ application by August 31st! 👉 more info evitasociety.org/index.php/atti… #oncosome #prostatecancer #evitians @CaterinaNardel1 @FrancescaBZNY

Many thanks to @CancerCell for the opportunity, to @FrancescaBZNY and @CaterinaNardel1 for the invaluable help and pristine writing, and @AIRC_it for the financial support.

In our research group at DemichelisLab (@FrancescaBZNY) we are currently working on plasma-derived EVs from prostate cancer patients: works like the one from @IcasanovaSalas are a treasure trove both for biomarker discovery/validation and for generating new hypotheses.

Casanova-Salas and colleagues analyze cfDNA, EV-DNA, and EV-RNA in prostate cancer longitudinal cohorts and identify signals of tumor adaptation to androgen receptor signaling inhibitors or taxanes treatment. doi.org/10.1016/j.ccel…

I had the honor of writing a preview of the article published by @ICasanovaSalas et al. in the latest issue of @Cancer_Cell . doi.org/10.1016/j.ccel…

@SimonDBarnett It makes sense (albeit the calculation should be done in 3d space). But it will be more lethal for any cell. The most important thing is to find specific targets. Does this proximity effect really matter when specificity allows to increase the dosage avoiding toxicity?

I'm currently bullish on radioligand therapies (RLTs) that target internalizing receptors. My guess is that internalized radiation is up to 5x more lethal per cell compared to radiation emitted from a cell's exterior. My napkin math is below. 🗒️⤵️ ~~~ RLTs kill by emitting high-energy particles. These particles travel in linear paths until they run out of energy, damaging everything in their wake. Radiation destroys cell organelles and turns oxygen atoms into volatile free-radicals, causing collateral damage. In ideal cases, these particles induce double-stranded DNA breaks (DSBs) within cancer cells. DSBs are hyper lethal to tumors. Released from the RLT, high-energy particles can travel in any direction. My hypothesis is that particles emitted from cell-internalized RLTs are more likely to impact the nucleus than those emitted at the cell's surface. This should increase the relative effect-per-RLT molecule, ceteris paribus. Left Image Math: 1. Assume that the cell and its nucleus are circular and that the nucleus is in the center of the cell. Assume that the nucleus is completely full of DNA. 2. R1 is the distance from the center of the cell to its membrane. R1 = 12.5 micrometers 3. R2 is the distance from the center of the nucleus to the nuclear membrane. R2 = 3.75 micrometers 4. Assume the RLT is bound to an infinitely small surface receptor, making it a point on the cell membrane. 5. The RLT sees the nucleus in its field-of-view (FOV). The fraction of its view obscured by the nucleus is θ while its full FOV is β. 6. The probability of a hit on the nucleus from a randomly emitted particle is P(hit), which is equal to the fraction of the full FOV taken up by the nucleus. P(hit) = θ/β 7. The angle θ subtended by the nucleus can be broken into two triangles. We can draw two, symmetrical right triangles with the shared hypotenuse R1. 8. This bisects θ, creating two equivalent angles (α). Using trig rules, we can get an expression for α, and thus θ as a function of R1 and R2. 9. Plugging 12.5 and 3.75 in for R1 and R2, respectively, we get roughly 0.1. 10. This suggests the hit probability of a randomly emitted particle from an RLT bound to the cell membrane is approximately 10%. Right Image Math: 1. If we assume, in the most extreme version of cell-internalization, that the RLT is located on the nuclear membrane. We can get an upper bound for the fold-improvement in probability from cell-internalized radiation. 2. If the RLT is a point on the nuclear membrane, there exists a tangent line to the nucleus. 3. By definition, the tangent line bisects the FOV of the RLT, creating the vertical dashed-line. 4. So, the subtended angle is half the total number of radians in a circle. P(hit) = α/2α = 1/2 5. This suggests the hit probability of a randomly emitted particle from an RLT bound to the nuclear membrane is 0.5 or 50%. ~~~ Therefore, there's a potential 5x probability improvement for cell-killing efficiency on the table for perfectly internalized RLTs. This is unrealistic, so I'm more comfortable saying there's a low-single-digit multiple improvement on the table. I made a lot of assumptions. I also worked this in 2D for simplification and I'm sure things would differ if I treated α as a 3D surface path in a spherical FOV. What's missing? What's too big an assumption? What is interesting to you in the RLT space?

@baym @jkpritch Clusters are made using PCA data. The UMAP is here just to show a nicer figure. I don't see any tremendous issue with this. But maybe I'm missing some relevant point of the story. Everyone looks like is going crazy with this UMAP!

To the point that @jkpritch and others have made, this seems like it has the strong potential to even further artificially inflate differences between clusters by giving the UMAP only the PCs chosen to most strongly separate the data

Also... the All of Us paper did a PCA on the matrix of SNPs, took the first 16 PC of a high-dimensional dataset and then did UMAP on that? What?

🥂🙌 A new doctor, Thomas Cantore, Ph.D. @cantore_thomas in the Computational and Functional Oncology DemichelisLab! #PRODE @CIBIO_UniTrento @paola_gasperini @YariCiani @MariannKruithof @NCIEytanRuppin

@5_utr @BrandonMeyersMD Are you skeptical about cfDNA in general?

@BrandonMeyersMD I am skeptical that it will add clinically actionable information beyond full and proper use of existing clinical information

I am beyond skeptical - Watch for touting the value of a new biomarker (🏦) that provides less information than basic clinical data and discarding/destruction of existing (free) clinical data

Happy 2024! #EVClub resumes with a study of #ExtracellularVesicles and #cfDNA to detect #HER2 positivity in #BreastCancer Need the login? Sign up: surveymonkey.com/r/EVClubISEV @IsevOrg @IsevComms @ISEV_Journals @LabWitwer @EVItaSociety isevjournals.onlinelibrary.wiley.com/doi/10.1002/je…

I am beyond excited to see our work out in @CD_AACR . We established a targeted sequencing assay to non-invasively monitor treatment resistance in advanced prostate cancer. Want to know more? Let’s take a short walk 👣 aacrjournals.org/cancerdiscover…

@dr_alphalyrae @klausenhauser On the other end, working with a bioinformatician that knows how the experiments are performed and understands biology, is a big plus.

@dr_alphalyrae @klausenhauser Coding and analyzing data needs concentration. Do you really want someone that does it in its spare time, while waiting for the end of a PCR or a western blot?

genuinely don’t understand the sharp line drawn across bioinformatics and molecular bio wet lab roles. When so many scientists do both, across industry (usually in startups) and also in academic labs