Articles
- Updates from the book, and the current state of their solutions. Practical and lots of nice graphs. Notes here.
Researchers are sitting on tech that could transform trees into power generators
- Project by team of students & researchers here, here
- Triboelectric effect of tree foliage: kinda like friction; leaves = positively charged -> produce small amts of electricity when contact neg charge things like tree trunks
- Can be harnessed for biological microgrid
- Genetically engineering elm tree -> tweak branch formation, thicker and denser leaves, pest repellance, faster growth
- Fully grown tree -> power 7 American houses! Alas, will take 40 years…
- Ethics of utilising something alive like this?
The Man Who Saw the Pandemic Coming
- Good read. Basically, human incursion into ecozones of other species due to road building, resource extraction -> closer contact with the huge pools of viruses out there. This was basically inevitable.
Nature Versus Nurture? Add ‘Noise’ to the Debate
- So relevant!
Crayfish
- All marbled crayfish -> derived from one individual— pet mutant female that could clone itself from eggs, got loose in 1990s. All genetically identical, so should be very vulnerable population…
- Despite lack of genetic diversity, lots of phenotypic diversity! Clones differ in color, size, behavior and longevity
- If not driven by nature (genes), then nurture (environment)? Still doesn’t explain all the variation we see. Recent research pinpoints intrinsic noise early in embryonic development!
Influence of noise
- “How much of who I am as a person is due to stochastic developmental events in my brain?” — Bassem Hassan, Paris Brain Institute
- Also from Hassan: “It’s the most difficult [factor] to actually control, to play with… You can play with the genome, you can play with the environment, you can play with physiology, you can activate certain cells and not others. … It’s a lot harder to manipulate the variation” and prove it to be the cause of differences in a trait of interest.
- New tools emerging to study/manipulate noise… esp in development, not just a factor systems deal with, but something they have evolved to take advantage of!
- Noise -> necessary driver of proper development, evolution. (like DsCam sets in Drosophila?)
- Michael Elowitz @ CalTech -> test variation in E. coli growing in same envt… transformed with two reporters with same regulation and either YFP or CFP -> cells expressed v different ratios!
- Cells make use of noise to create necessary variability in their behavior and biological state.
- Does this variation balance out at the multicellular level? Difficult to answer, but probably not- ex. different behavior of the crawfish. But now…
Armadillo
- Nine-banded armadillos always have litters of identical quadruplets -> great system to study role of noise in development (Jesse Gillis, CSHL)
- Classic random event— X-chromosome inactivation—begin as early on as 25 cell embryos! (Each armadillo thus has precise and random combos of 25 maternal or paternal X chrom selections). Huge downstream effects bc so early.
- Looked at other 31 chromosome pairs— not completely active/silenced, but still some determination in how active the maternal or paternal chrom is happens in development… used ML to see when the ratios of their contrib to gene expression stabilised— at stage of hundreds of cells! Again, very early, exponential effect over time…
- Estimated that developmental noise -> 10% of variation among the litter!!!!
- (Studies in neural development of flies -> at least 30-45% variation in walking behavior attrib to random diff in wiring patterns.)
Research Readings
Deep learning: new computational modelling techniques for genomics
- Highlighted PDF
Chromatin accessibility dynamics in a model of human forebrain development
- Rationale: Development of forebrain, incl expanded human cerebral cortex, is difficult to study — lack of models, inaccessible to study or manipulate… tracking epigenetic changes, lineage tracing over time has lots of potential to reveal molecular mechanisms underlying normal development and disease states
- Approach: Used human induced pluripotent stem cell (hiPS) cell–derived three-dimensional (3D) organoids to make models of specific forebrain regions (cortical and subpallial); cultured for 20 months! Took samples for ATAC-seq, RNA-seq at different time points. Validated lineage tracing with primary human tissue samples.
- Results: Three patterns of chromatin accessibility across samples (diff in vitro models and clinical ones at diff time points); also compared clinical vs hSC-derived accessibility over time— later time points got more similar to each other
- To compare in vitro model to cortical dev over time, used ENCODE and other recent primary epigenetic datasets— computed Jaccard similarity at diff time points… overall: forebrain organoids undergo chromatin remodelling similar to primary tissue; consistent across multiple data types and long time-scales (early fetal to postnatal!)
- Compared patterns of chromatin accessibility among lineages from human fetal tissue (HFT) and organoids: ID’ed 6 clusters for diff time points/lineages
- Next associated gene expression programs with this landscape— looked for enhancer-gene linkages based on “coordination of accessibility and gene expression”
- Cx with disease signatures? Mapped SNPs associated with disease risk from GWAS onto landscape, used linkage disequilibrium score regression -> enrichment of risk variants in each cluster; certain clusters enriched with variants for schizophrenia, autism spectrum disorders, ADHD
- How evolutionarily conservation varies across data? Used PhyloP scores (lower = less conservation), which were different across the clusters. Also can measure conservation in vivo with enhancer driven reporters; quantified overlap with VISTA mouse dataset— don’t quite understand these analyses.
- Found enrichment of hARs (= human accelerated regions, non coding genomic loci involved in brain dev with high vertebrate conservation & greater substitution rates in humans vs other great apes) across clusters
- Sequence features in ATAC-seq data: Enrichment DNA sequence motifs in clusters (found, ex., motifs regulating pluripotency in early lineages); used chromVAR and RNA-seq to find more unbiased links with specific TFs; verified some finds experimentally (ex. NFIA motif accessibility corr with NFIA expression in glial lineage, esp progenitors, and NFIB… other trends suggested other TFs related to lineage differentiation in forebrain development)
- “To explore the diversity of sequence motifs in each of our dynamic clusters, we grouped enriched motif according to their JASPAR family assignments and computed the Shannon entropy of these labels by using the total number of enriched motifs as a prior estimate” -> period from 79-230 days assoc with most diverse TF activity during cortical neurogenesis
- Applied whole framework to analyze important TFs in corticoneurogenesis (sequential gen of layer-specific neurons over 20 wks), with motif analysis, chromVAR, gene expression. Diff “waves” of TF activity; new role of MEF2C in midcortical neurogenesis (in addition to validating TBR1 with early neurons, BRN2 with late)
- All three assoc with ASD -> downstream fx and pathways in corticogenesis? Target genes converged on signature of axonogenesis and chemotaxis
- Significance: Shows overall utility and function of cortical organoid models & analysis pipeline; power of integrating chromatin accessibility & info on regulatory elements with gene expression to get insight into differentiation & fate specification; ability to link disease risk to neurogenesis and gliogenesis
Hypoxia and Innate Immunity: Keeping Up with the HIFsters
- How HIF stabilisation supports innate immunity
- Intx btw immune system & metabolism -> oxygen is important resource; chronic inflammation (of protective and pathological varieties) = hypoxic -> HIF regulation and responses are involved
- Highlighted PDF saved elsewhere!
- Motivation: Major challenge = predicting quant effects of gene reg elements (GREs) on expression— difficult to generate necessary large datasets!
- Approach: ( = uASPIre: ultradeep Acquisition of Sequence-Phenotype Interrelations) Created system of three components— 1) a GRE, 2) recombinase w expression controlled by the GRE, 3) DNA substrate that can be modified by recombinase
- Each #3 -> info about GRE sequence and measure of its function; can be read out with NGS; can assess large libraries of GREs with same principle. Also can increase resolution of readout simply by upping copy number.
- For #2, used bacteriophage integrate Bxb1 (the one used to encode “trits” in that other paper!), causes irreversible inversion between recognition sites (in this case, flips mCherry to sense-orientation)
- Validation: Recorded translation kinetics of 300k bacterial RBS -> 2.7 million seq-expression pair
- Replaced mCherry from readout piece above w 150 bp non-coding DNA; transfect bacteria with library of RBSs (part of 5’ UTR of mRNA, so right after promoter of Bxb1); induce, sample over time, extract plasmids, and NGS
- Not subject to bias from PCR amplification!
- Generalized from dataset w new deep learning approach (ensemble + uncertainty modeling, uses ensemble of 10 residual CNNs, = SAPIENs: Sequence- Activity Prediction In Ensemble of Networks) predicting fx of untested RBS’s with high accuracy; eliminated systematic biases of other approaches; no plateau reached even at highest training set size -> even more potential
- Applied integrated gradients to pull interpretation from model & found legit features; in silico evolution to design new RBSs worked too (basically found Shine-Delgado seq from nothing)
- Whole pipeline can be applied to other sorts of features as well, not just RBS!
Long-range single-molecule mapping of chromatin accessibility in eukaryotes
- (Only the abstract, no access atm)
- Mapping open chromatin -> identity regulatory regions; however, current methods rely on DNA fragmentation & short reads -> can’t really get large-scale info or coordination between cell states
- Here, develop single-molecule long-read accessible chromatin seq (SMAC-seq)— profiles individual chromatin fibers; get high-res, multi-kb assessment of chromatin states
- Use EcoGII, a methyltransferase with low seq specificity to pref methylate open chromatin + nanopore seq to read out DNA modifications
- Validation: prove they match bulk accessibility measurements, get single-molecule nucleosome and TF protection footprints (!!), quant correlation btw chromatin states of distal regulatory elements (aka enhancers?)
Imaging cell lineage with a synthetic digital recording system
- intMEMOIR!
- Would be nice to engineer cells to actively record lineage that could be read out in situ -> use serine-integrase based recording system, validate ability to lineage trace in cultured stem cells and flies
- System = array of independent three-state genetic elements that can recombine stochastically + independently -> ~60k digital states; image-based read out with smFISH along with endogenous transcripts -> no need for seq
- Use of “trits” > “bits” yields more flexibility, memory, stability of info.
- Used phage serine integrases to record in trits: irreversibly mediate recomb between directional attP and attB sites, inverting or deleting the seq depending on orientation; don’t rely on DNA repair mechanisms; work across species.
- Each trit = flanked by att sites so that it was either inverted or deleted or kept the same; driven by strong promoter to enable smFISH read out
- Combined a bunch of these (1 of 10 possible dinuc flanked by att sites = a trit), integrated into safe harbour locus. Doesn’t have same downsides as gene editing strategies of barcode collapse!
- Validated concept, then developed cell line from mESCs (intMEM1) for inducible, automatic recording (dox-inducible expression of integrase Bxb1, 10-trit array)
- Can read out whole array with 5 rounds imaging, 4 fluorophores, 20 probes
- Continued with lots of validation; quant editing rates, ability to reconstruct lineages vs ground truth
- Looked at spatial organization of clones, gene expression, development in fly brains!
- Don’t have access to full article atm, but in summary…
- Hard to synthesise gRNA libraries, screens limited sometimes by lack of diversity
- Developed CRISPR adaptation-mediated library manufacturing (CALM) using hyperactive S. pyrogenes CRISPR-Cas acquisition machinery (Cas1, Cas2 abs all) in S. aureus bacteria
- Introduce human DNA of interest to bacteria -> cells act as factories, making 100ks of crRNAs targeting much of the human genome (100 guides per average gene!) Can iterate to get dual-gene libraries!
- Somehow, the “polarised nature of spacer acquisition” reveal “historical contingency” relating to antibiotic resistance? IDK, this sentence (and the title) did not make sense to me. Cool idea overall though!
- From Octant!
- Humans have 813 GPCRs, super important for signalling; top drug target! But hard to study signal transduction and structure
- Here: platform to characterize large libraries of variants in human cells w barcodes transcriptional reporter; tested almost all 7.8k single a.a. variants ( = deep mutational scanning, DMS) of B2AR (beta-2 adrenergic receptor) at 4 concentrations of agonist, measure response through cAMP pathway ->
- ID residues important for signaling & modulation, that may lead to loss of function from known human mutations
- Unsupervised learning -> resolve all residues crucial to signaling, known conserved structures, a new structure conserved across Class A GPCRs!
- Reporter: B2AR signals to Gs protein -> activates AC -> cAMP production -> stimulate transcription of barcoded reporter gene thru CRE (cAMP responsive element!) -> read out with RNA-seq. Designed vector with lots of fancy stuff to integrate stably, not cause toxicity, good SNR; also knocked out endogenous B2AR in validation cell lines. Showed dose dependent response of barcode readout to antagonist! “Massively-paralleled” bc incorporated 15 bp oligo into 3’ UTR region of each mutant when constructing library, so you can pool everything, don’t need sc seq
-
Once tested, developed variant-activity landscape heat map (looks very cool!). Could see s different mutation subtypes (ex. effects on different TM domains)
- Population genetics: compared results with Genome Aggregation Database- individual mutations/variants, which could guide future characterization of mutants; more robust analysis possible by aggregating results from mutations at a single site
- Unsupervised learning: used UMAP, hierarchical clustering to functionally cluster mutations (tolerant/intolerant?), map back to structure!
Tumour heterogeneity and the evolutionary trade-offs of cancer
- Review of theory of multi-task evolution of cancer: tumors evolve in host, face selection trade-offs btw biological functions. Distinguishes these diff biological tasks, “specialist” vs “generalist” tumors & relation to driver mutations. Can help explain intratumor heterogeneity, spatial organization of tumors.
- Important to have theoretical framework (here of evolutionary reasoning) in addition to all the data on heterogeneity! Ex: Tumor seq data + principles of population genetics + formalisation of nature of selection -> explain genetic alterations, clonal architectures & reveal phenotypes being selected for— also how trade-offs lead to diversity (“go or grow”, prolif vs survival
- Includes some overview of ParTI algorithm, which can take bulk or scRNA-seq to analyze for archetypes (specialized tasks), and use clinical markers or network analysis to determine what tasks each archetype may correspond to
- Theory can also be used to find vectors of driver mutations (vs passenger) that were selected for to move cells closer to Pareto front (somewhere on the polyhedron of archetypes)
- Related to transcriptional noise project (second hypothesis): selection occurring at total tissue level, where trade-offs & tumor-level selection contribute to division of labor, which maximizes total tissue performance
- [Article fully highlighted on Books app]
Mitochondrial distress call moves to the cytosol to trigger a response to stress
- How do mitochondria send signals of stress through cytosol to nucleus? In mammals, via integrated stress response (ISR): overall decrease in protein production, increase in some TFs. But how is signal transduced?
- Ex. ISR mediated by phosphor of eIF2a- 4 kinases for 4 possible sites for 4 diff reasons
- Two groups (including Kampmann’s, with genome-wide CRISPRi screen) simultaneously identifies upstream pathway of one of the phosphorylation events (necessary and sufficient) using similar approaches in mammalian cells: one used fluorescent CHOP as signal of ISR induction & introduced random mutations to find genes triggering it; one used fluorescent AT4 (CHOP & AT4 both TFs activates by eIF2a
- Basically: protein OMA1 on inner mitochondrial membrane, when depolarised, cleaves part of little-studied DELE1, which enters cytosol and binds to HRI, which phosphorylates eIF2a. Both showed accumulation of cleaved DELE1 portion in cytoplasm alone could trigger ISR; also that HRI could be activated even without heme depletion (previously unknown)
Visualizing the genome in high resolution challenges our textbook understanding
- Basically, chromatin organization is complicated and dynamic and we are still tryna understand it!
- Super-resolution imaging (STORM, Oligopaint), CRISPR labelling, single-molecule tracking in live cells… all helping elucidate at higher resolutions, dynamics of different processes
- Whole idea that actual droplet & phase separation is significant in regulation of chromatin and transcription
- See highlighted article in database
A Deep Learning Approach to Antibiotic Discovery
- (Link actually to Quanta article)
- Taking a hypothesis GENERATING vs TESTING approach! to drug discovery: letting ML discover structural features of compounds that are effective antibiotics, rather than conventional approach of looking for things that target specific processes of bacterial survival (ex. protein synthesis, cell wall, etc.).
- Reveals previously non-antibiotic related small molecule drugs that disrupt unexpected pathways (example here was electrochemical gradients)
- Important to remember it’s not AI saving the world… but an algorithm helping humans figure things out, still relies on experimental validation, good test/training data
The causes of evolvability and their evolution
Evolvability
= ability of a biological system to produce phenotypic variation that is heritable and adaptive_
- Heritable phenotypic variation = raw material of evolution
- Potential to produce adaptive (beneficial) variation = essential for natural selection
- Here, focus on experimental evidence/research & three major causes of pre-mutation evolvability
- Also: Can evolvability evolve? 1. Is there genetic variation in it? (YES) 2. Does it actually evolve in nature/the lab? (Generally, YES) 3. Is a change in evolvability adaptive or the byproduct of other processes? (OPEN QUESTION)
- Another open question— how do causes of evolvability interact?
1. Phenotypic heterogeneity
= genetic & non-genetic mechanisms that create phenotypic het that can become heritable by epigenetic, modification or mutation (focus here on non-genetic)
- Four mechanisms:
- Stochastic gene expression AKA gene expression noise
- Contribute to diversity in phenotypes like viral latency, bacterial competence, antibiotic/cancer drug resistance….
- Persister phenotype — some cells of isogenic pop’n become dormant -> survive drugs; pop’n bottleneck from selection does not actually limit “supply” of beneficial mutations as might be expected. (Mutliple evolutionary paths to resistance shown to remain)
- Rare cell viability — subpop’n stochastically expresses phenotype to avoid drug (but not dormant); can occur before tx -> “pre-resistant” to drug exposure
- Facilitates genetic assimilation = new phenotype from environmental perturbation becomes genetically encoded
- Can help evolvability by changing how strongly mutations affect fitness- ::stronger positive effects of beneficial mutations with more heterogeneity?::
- Errors in protein synthesis AKA phenotypic mutations
- Create variation in protein pool from one gene, some of which may be adaptive; pave way for adaptive genetic change
- Ex. Stop-codon readthrough
- Epigenetic modifications
- Prions: Ex. aggregation reduces translational fidelity -> errors like frameshift, stop-codon read through -> reveal cryptic genetic variation (normally causes little phenotypic variation, but can in new environments or genetic backgrounds)
- DNA and histone methylation: Heritable modifications that can create variation (Ex. intratumor— modifications to reg elements -> heritable variation in chromatin structure)
- Protein promiscuity
- = Proteins that have one primary fx and other secondary ones (‘moonlighting’ catalytic activities); provide reservoir of activities that can be enhanced by gene duplication -> mutations creaiting two new proteins
- Also applies to promiscuous activity of reg elements (eg. enhancers)
- Does it evolve? Is there variation? Yes! Ex. Heterogeneity in DNA repair machinery -> changes to mutation rate; gene expression noise varies with promoter strength; seems to actually be adaptive in some cases
2. Robustness
= ability of population to explore new genotypes without disrupting essential phenotypes, lack of phenotypic variation in presence of genetic variation
- Increase evolvability by…
- Facilitating cryptic genetic variation via, eg. .DNA mutations that enhance protein stability; gene duplication; chaperone proteins that protect against misfiling
- Revealing diversity -> new genetic backgrounds for mutations to manifest, -> diff phenotypic effects via epistatic intx (non-additive intx between alleles contribution to a phenotype)
- Incr access to diverse mutational neighborhoods
- Examples of robustness -> evolvability
- Enabled expansion & diversification of C2H2 zinc fingers in metazoans!
- In reg elements (ex. TFs can bind to many mutationally-related sequences) -> new binding sites, testing mutations
- In gene expression, allows exploration of gene reg networks + sampling new config while preserving ancestral roles of protein -> new networks, fxs
- Research helped by new tech like deep mutational scanning, ancestral protein reconstruction
- Does it evolve? Is there variation? Yes, via increased protein stability, gene duplication
3. Topological features of adaptive landscape
= analogous to physical map where coordinate is point in genotype space, elevation is fitness or other phenotypic readout
- Landscape & population’s location within it -> amount of adaptive phenotypic variation mutations can bring forth
- Shape of peaks dictated by epistasis
- Positive epistasis = synergistic; convex, smooth peaks, good for evolvability
- Negative epistasis = antagonistic (beneficial effects cancel out); concave peaks; bad for evolvability
- Sign epistasis = sign (beneficial or detrimental) of double mutation is diff than one or both of the individual ones -> local peaks and valleys, ruggedness — bad for evolvability
- Mapping facilitated by experimental evolution, engineering all Its in a specific genotype space, deep mutational scanning (assay phenotypes for many mutations of single WT genotype)
- TF binding sites = challenging genotype space to explore — fewer dimensions -> lower evolvability (higher dimensions allow ‘extradimensional bypassing’ of saddle pts in rugged landscapes
- Address with things like Aviv Regev ML paper on testing every single 80mer as promoter?
- Does it evolve? Is there variation? Yes; location of pop’n or individual on landscape changes with mutations & recombination; evolvability conferred by local topography can evolve (from long-term experimental evolution studies with E. coli)
Synthetic lethality as an engine for cancer drug target discovery
Intro
- First wave of target cancer therapies -> drug recurrently mutated genes, but approach being exhausted — remaining genetic drivers = ‘undraggable’ by structure, result in fx loss
- Need combo regimes for more long term results; targeting more than just genetic factors
- Immunotherapy (esp PDL1) shows some promise, but still limited
Synthetic lethality
= when inactivation of one of a pair of genes individually doesn’t do much, but both is deadly (in cancer, one could be mutated & the other target by drug -> selectively kill cancer cells)
- Ex. BRCA1 or BRCA2 and PARP = synthetically lethal pairs, PARP inhib for BRCA-mutant ovarian cancer = first clinical ex of synthetic lethally
Methods to discover new targets
- Advent of CRISPR/Cas9 fx genomics -> new approaches to discover targets
- Includes details on how to decide what CRISPR gRNA library to use
- Many large-scale gene knockout screens -> many targets remain undiscovered; however, more efficient methods at target ID (esp for combinations) necessary
- Critical to consider genetic context to avoid signal dilution bc heterogeneity in clinical trials & exp design
- = Driver mutation, tissue of origin, other fx genetic lesions in a subgroup of cancers
- Ex. selective dependency of certain cancer lines on O2 sensor EGLN1… basically, it’s complex
- Looking for unmarked oncogenes (not genetically activated thru mutation, amplification, or translocation, but still important drivers in certain genetic contexts) with fx genomics screens
- Approach:
- Anchor screen: uncover context-dependent unmarked oncogenes with CRISPR library in combo with a target drug (anchor) relevant to specific genetic context
- Discover synthetic lethal approaches that don’t necessary kill cancer cell, but promote immune response: tumor intrinsic immune invasion targets
- 1) Identify genetic context conferring immune evasion - by modified CRISPR screen with diff readout?
- 2) Identify drug targets that reverse phenotype
Targeting tumor suppressor gene loss
- Seemingly undraggable bc function, gene itself are sometimes gone -> finding druggable synthetic lethal partner = only approach; feasible only by fx genomics screen (rather than hypothesis-driven) since so many potential intx
- Ex. PRMT5 dependence in cells with MTAP deletions
- Subset of synth lethality — collateral lethality: ‘passenger’ gene to tumor suppressor lost in addition to this driver (here, MTAP often deleted alongside actual tumor suppressor, CDKN2A) -> implicated in mechanism of dependence on PRMT5
Dissecting cellular crosstalk by sequencing physically interacting cells
- PIC-seq — notes on PDF
- Rewind — notes on PDF, but so cool!
- Barcode w/ lenti library initial cell population -> expand -> split into two “carbon copies”
- Set 1: treat with drug, select for resistant cells, sequence to get barcode, make RNA FISH probes for resistant cell barcodes
- Set 2: Label with your FISH probes to tag cells PRE-selection/differentiation, sort, and sequence!
3D ATAC-PALM: super-resolution imaging of the accessible genome
- Goal: visualise 3D architecture of accessible genome at nanometer scale (superresolution) in situ
- Approach: integrate ATAC (assay for transposase-accessible chromatin) with visualization via PALM super-resolution imaging and lattice light-sheet microscopy
- Multiplex with oligopaint DNA FISH, RNA FISH, protein fluorescence -> connects microscopy + genomic data
- Reveals spatially-segregated accessible chromatin domains (ACDs) enclosing active chromatin, transcribed genes
- Showed that CTCF prevents clustering of accessible chromatin, decompacts ACDs
- Mechanism: conjugate transposase with photoactivable fluor… lots of very complicated microscopy things
- Amazing! “To quantify the degree of accessible chromatin clustering, we used the pair autocorrelation function g(r), which was applied by physicists to describe the clustering of stars in galaxies.”
Astrocyte layers in the mammalian cerebral cortexvrevealed by a single-cell in situ transcriptomic map #science/neuro #science/singlecell
- Notes on PDF
A transcriptomic atlas of the mouse cerebellum reveals regional specializations and novel cell types
- Motivation: Cerebellum is important; inventory of cell types in it is lacking!
- Composed of one circuit repeated thousands of times: mossy fibers from many brain regions excite granule cells (GCs) excite Purkinje cells (PCs) — sole outputs of the cerebellar cortex
- Other circuit elements = inhib neurons (molecular layer interneurons — MLIs), Golgi cells — GCs), etc. -> contribute to more complexity, regional specialisation?
- Challenges to survey of cell types: mostly GCs, so hard to sample rarer subtypes; existing molecular char of cell types is limited;
- Approach: high-throughout single nucleus RNA-seq (snRNA-seq), clustered transcriptomes -> 46 clusters and mapped each to one of 18 known cell types; then explored inter-cluster heterogeneity within those to see if there were undiscovered thing
- Electrophysiological studies by, ex, applying glutamate and observing spike responses
- Pipeline “to record from individual MLIs in brain slices, image their morphologies, and then ascertain their molecular MLI1/MLI2 identities by smFISH” — link function, morphology, gene expression
- Looked at spatial distribution & variation of cell types, gene expression in different globules
- Results:
- Regional specialisation of Purkinje (PCs), granule neurons— found 9 PC clusters corresponding to diff globules based on Aldoc expression
- Continuous transcriptional variation for several interneuron subtypes— surprising for unipolar brush cells (UBCs), had spectrum of electrophys proprieties & gene expression
- Surprised by two types of molecular layer interneurons (MLIs)! Continuous morph differences, but v distinct electrophys behavior
A systems approach to infectious disease
- Framework for a systems biology approach to any question, really, with a few extra considerations re: infectious diseases. Notes on PDF
Nongenetic Mechanisms of Drug Resistance in Melanoma
- Relatively self-explanatory… highlighted PDF
Podcasts
- I love my interviews with super random people that I’ve never heard of, and who generally end up being really cool. This one’s with a (quite successful) PhD student studying math at MIT… you know, after having played a few years for the Ravens in the NFL!
Ezra Klein with Rebecca Solnit
- Such a lovely discussion between such deep-thinking, contemplative people. Such a joy to listen to, and I learn so many new ways to think about the world— and ways to think about how to articulate my own views.
(Unedited) Nicholas Christakis with Krista Tippet
- DNF
Books
The Truth About Animals — Lucy Cooke — 3/6/20
- This one grew on me. Cooke presents a series of histories of chronically “misunderstood” animals, from sloths to moose to penguins. Some of the myths she cites are hilariously absurd, some linked to darker & disturbing experiments, some downright raunchy— and the fun facts & actual wild truths she conveys are just as fantastic!
- Plus, I can never resist a “narrated by the author” :)
Horizon — Barry Lopez — 3/9/20
- Beautiful, beautiful. I treasured this one slowly, in the way you can only really do with a hardcover book. It was a meditation, an absolute pleasure. Notes are a collection of new words that piqued my interest and/or sounded pretty, & passages that did the same— so many evocative scenes of nature, color, pattern.
Meander, Spiral, Explode — Jane Alison — 3/13/20
- Different from my usual, but a refreshing dive into the world of narratives, the relationships between visual pattern & design and how we move through & experience stories.
- Definitely is/will change the way I think about the narratives I read and write; the natural inspiration of structures that underlay and drive a story
- Would be cool to look at structure of episodes & seasons of /This Is Us/ with this perspective
Searching for Stars on an Island in Maine — Alan Lightman — 3/18/29
- A beautiful exploration of faith & science, of permanence & unity & certainty, of what it means to be human. Read the real book.
Cancerland: A Medical Memoir — David Scadden, with Michael d’Antonio — 3/21/20
- Interesting insights on how Asylimar conference and temporary ban on genetic engineering early on in Boston actually set them on the path to being ahead of the game— had a plan, process to negotiate doing science with public concern
- Cancer + aging linked because both involve changes in regulation of cell growth and survival; the latter with senescence, but this is countered by inflammation; throws of balance of immune system and makes older ppl more vulnerable to cancer, other infectious diseases
- OTOH, can cancer’s ability to regenerate/replicate on end be used for regeneration, healing, anti-aging therapies? Vice versa, can mechanisms of senescence be applied to combat uncontrolled cancer growth?
- The /just do something/ impulse that drives so much or oncology care (by patients & doctors), even when it’s just prolonging death — made worse by FDA approval of drugs that induce remissions but don’t actually extend survival beyond a few months.
- Founded Harvard Stem Cell Institute— in response to big challenges in funding for stem cell research since govt ceased funding; public controversy
No One Is Too Small To Make A Difference — Greta Thunberg — 3/22/20
- Thunberg’s collection of speeches offer an honest, direct, and personal perspective. She makes her point strikingly, effectively, rationally, and human-ly, without fear of assigning justified blame. Yet she just as cogently conveys the simple optimism of our agency to address the climate crisis: the solutions are known, all we must do is change.
- Honestly, can never read/listen enough about climate change & why we need to get our shit together. I & we need every reminder we can get.
- /Cathedral thinking/— sometimes we must start laying the foundation before we know what the ceiling will look like
The Philosophy of Loyalty — Josiah Royce — 3/22/19
- This one’s been a long time coming. A series of lecture mentioned by Atul Gawande in On Being; read the first half last summer and powered through the rest in the past few days. Notes highlighted on Google Play, but basically, introduces a philosophy with loyalty as man’s chief moral good
Three Women — Lisa Taddeo — 3/24/20
- This was… different. A revolutionary and extremely well-written piece of journalism; I just was not quite in the mood/state of mind to appreciate it. Read if you dare. feel the need; I probably should not have pursued this one.
The Jungle Grows Back — Robert Kagan — 3/27/20
- America and Our Imperiled World
- The liberal world order (post-WWII “peace” led by US et al) = /fragile & impermanent/; like a garden under seize from the forces of a jungle… it’s not normal, given context of human history!
- Should the US continue carrying such a moral/social “burden” and interventionist attitude? Or try to be a “normal” country?
- History, history, history
- European democracies are mostly young and untested, protected by liberal workd order… as that changes, can only suspect more surges of far-right nationalism, facism…
- Criticism from both the left & right that the “liberal world order” is really just America asserting its hegemony
- To be quite honest, I was zoned out for the vast majority of this.
The Strange Order of Things — Antonio Damasio — 3/29/20
- Feelings, consciousness, neuroscience,… very philosophy, very meta, a satisfying framework that shows just how intertwined and the /same/ our biological and felt identities are.