Pall Melsted Profile Banner
Pall Melsted Profile
Pall Melsted

@pmelsted

2,623
Followers
319
Following
247
Media
3,764
Statuses

Professor @hi .is in CS, head of rna seq data analysis at decode genetics (views are mine). Bioinformatician, epistemic trespasser, &c. ps. I hate GTF files

Joined January 2013
Don't wanna be here? Send us removal request.
@pmelsted
Pall Melsted
6 months
In this preprint with @sindri_e we compared seven widely used methods for batch correction of single cell RNA-seq data. We found that all but one of the methods introduce batch effects when there are none. 1/N
5
138
474
@pmelsted
Pall Melsted
6 years
I'm pretty sure my morning coffee changes my gene expression by more than 7%
5
88
483
@pmelsted
Pall Melsted
4 years
Exactly one month ago Iceland had 1045 active diagnosed infections. Today we have 99 active, and added 608 new cases during this time, most of which have recovered. #TestTraceIsolate
Tweet media one
5
59
214
@pmelsted
Pall Melsted
6 years
Today I am doing proper bioinformatics. Matching up gene symbols between databases
11
15
207
@pmelsted
Pall Melsted
4 years
A twitter play with three acts
Tweet media one
Tweet media two
Tweet media three
2
34
201
@pmelsted
Pall Melsted
6 years
Next time your kid loses a tooth, put it in a coke-bath and have them watch what happens. There is an important lesson to be learned 1/n
Tweet media one
2
57
142
@pmelsted
Pall Melsted
7 years
New release of kallisto is out kallisto can now produce BAM files in genomic coordinates, sorted and indexed for IGV consumption, details in a new blogpost Oh, and it runs on a tiny computer
3
85
136
@pmelsted
Pall Melsted
6 years
Society will teach your kids about dental hygiene, dangers of soda drinks, but who is going to talk to them about experimental design? 4/4
4
19
132
@pmelsted
Pall Melsted
4 years
Speed is crucial in fighting this epidemic. In Iceland we use test-trace-isolate to curb the spread in addition to restrictions on social gathering, 2m social distancing and mask wearing in public. Here is a rough timeline of what happens for a positive diagnosis
3
43
125
@pmelsted
Pall Melsted
5 years
I can solve every problem with a combination of cut,grep,sed,sort,uniq, and awk
5
13
123
@pmelsted
Pall Melsted
7 years
Transcripts are real, genes are a useful abstraction #gi2017
4
45
114
@pmelsted
Pall Melsted
7 years
I am a real bioinformagician, I can do analysis stuff without first converting data into tabular form
8
16
111
@pmelsted
Pall Melsted
6 years
New preprint out w @lpachter & @vntranos on the Barcode, UMI, Set format (BUS) for processing of single cell RNA-Seq datasets. Shows how to create modular frameworks for processing scRNA-Seq data, fast using kallisto (v 0.45.0 just released) and simple
2
46
109
@pmelsted
Pall Melsted
3 years
Recursive vaccinations! My brother got his first and only J&J/Jansen shot today. Yes, that’s a tattoo of himself
Tweet media one
2
11
102
@pmelsted
Pall Melsted
4 years
deCODE genetics is has stepped up to assist with covid19 screening in Iceland. Started last Friday and have completed 1800 tests and are operating at a rate of 1000 tests per day. That’s 0.3% of the population per day.
3
32
98
@pmelsted
Pall Melsted
4 years
Border screening in Iceland has detected 13 cases of the British variant. Twelve came from the UK, one from Denmark. Every positive sample is sequenced.
0
20
93
@pmelsted
Pall Melsted
6 years
Break it to them gently that their paper got rejected because reviewer 3 had serious concerns about lack of negative control 3/4
2
10
93
@pmelsted
Pall Melsted
6 years
Gene databases: so there is this gene called LAMP3 Me: sounds good Gene db: but wait there is also CD63 Me: ok? Gene db: and it’s also known as LAMP-3, I mean what could go wrong Me: [...]
11
19
89
@pmelsted
Pall Melsted
8 years
Good news: I have tenure. Bad news: I'm the new department head.
12
8
89
@pmelsted
Pall Melsted
3 years
Iceland has found 12 cases of omicron confirmed with sequencing. All PCR positive cases are sequenced. Since the first report of omicron, all S dropouts have been sequenced within 24 hours. The US has confirmed 20 cases, how large is the the iceberg?
7
15
86
@pmelsted
Pall Melsted
5 years
Yesterday was my 10 year anniversary in bioinformatics. Thanks to @jkpritch for taking a chance on me when I knew nothing about genomics, sequencing, and rna. Heading out to #ASHG19 now
1
2
86
@pmelsted
Pall Melsted
6 months
The paper is the result of this thread pulling. My advice is twofold and simple. 1. just use Harmony 2. if you are correcting for batch effects, make sure it does nothing when there are no batch effects. 9/N (N=9)
3
10
87
@pmelsted
Pall Melsted
4 years
Every polynomial has their day
@JSEllenberg
Jordan Ellenberg
4 years
Well, I'll be damned, it really is a cubic.
Tweet media one
200
3K
15K
1
8
84
@pmelsted
Pall Melsted
3 years
Triple blind review. Authors don’t know the identity of reviewers and vice versa. Reviewers don’t know the identity of the journal. Should the quality of review and rigor depend on where it was submitted?
8
4
75
@pmelsted
Pall Melsted
4 years
The paper from @decodegenetics on the spread of SARS-CoV-2 in the Icelandic population is finally out proud to have worked on this with @hakon_jon and others
3
32
74
@pmelsted
Pall Melsted
8 years
. @NatureNews you know the image is of JavaScript code, right?
1
15
71
@pmelsted
Pall Melsted
5 years
Bifrost – Highly parallel construction and indexing of colored and compacted de Bruijn graphs. Paper by @GuillaumOleSan and myself, constructs colored de Bruijn graphs for 118K Salmonella strains. Software: , preprint:
1
24
70
@pmelsted
Pall Melsted
9 years
uniq on mac only compares first 8kb of line, use sort -u instead #waytoospecificcmdlineadvice
12
59
68
@pmelsted
Pall Melsted
3 years
I’m psyched about the vaccine lottery this week in Iceland. No $1M prize, we will randomly draw a birth year (1975-2005) and all born on that year will be vaccinated that time. And we’ll keep going till we’re done. 🤞1980
3
4
65
@pmelsted
Pall Melsted
4 years
Since we are cleaning up gene names for excel, how about we don’t put a slash in the name. Pretty please
Tweet media one
3
6
62
@pmelsted
Pall Melsted
8 years
Kallisto is published. I've had some straight-to-video papers, this is not one of them
4
62
59
@pmelsted
Pall Melsted
6 years
My phone has learned how to spell ensembl, pseudoalignment and Rnaseq. Ducking amazing #DeepLearning
4
3
58
@pmelsted
Pall Melsted
2 years
Genome informatics 2022 will be held as a hybrid conference in Hinxton, UK Sep 21-23 this year. Right now you have one week to submit your abstract for #GI2022 (deadline July 12th) You know what to do.
1
20
55
@pmelsted
Pall Melsted
6 years
We have a position available at the University of Iceland building up bioinformatics services as well as working on bioinformatics in my group. deadline is april 18th, plz retweet.
7
109
52
@pmelsted
Pall Melsted
4 years
Yesterday Iceland received 10K doses of the Pfizer vaccine (for 5K people). Today 2.5K were vaccinated and we will finish the doses tomorrow. Throughput is estimated to be at least 10K per day. Scaled up to the US this corresponds to 2.5M people.
3
10
52
@pmelsted
Pall Melsted
3 years
New preprint out, lead by @kreldjarn & @SolviStats Two key results. 1. Reconstruction of a giant infection tree with 2500+ infected individuals from the "third wave" of covid in Iceland. 2. Using this tree to simulate the effect of vaccinations 1/5
1
18
52
@pmelsted
Pall Melsted
9 months
Genome Informatics 2023 is about to start #gi2023
Tweet media one
2
5
50
@pmelsted
Pall Melsted
9 years
I don't always analyze RNA-Seq in the cloud. But when I do, I'm at 30k ft on my laptop http://t.co/wImr7XBalJ #bog15 http://t.co/BLQtBBedkh
Tweet media one
3
18
48
@pmelsted
Pall Melsted
9 years
If you are struggling with understanding BWT on graphs, look at interactive viz from my student @Moyaccercchi
3
33
51
@pmelsted
Pall Melsted
3 years
I’ve written software that eats millions of reads in a matter of 10 seconds. I don’t hesitate grepping huge files and feeding them into a series of pipes of awk/cut/sort/uniq that I write with my eyes closed. But canvas takes 10 seconds to list discussion items 🤷
2
1
50
@pmelsted
Pall Melsted
6 years
Hate is a strong word, but I really hate GTF files
5
7
47
@pmelsted
Pall Melsted
5 years
My brains feels like mush. It’s either because I’m 40 today or the 8 hour jet lag, can’t tell because batch effects
5
0
49
@pmelsted
Pall Melsted
6 years
Setting up new computer: 0 min, where did all the brew packages go? 2 min, what is this bioconda thing I keep hearing of 5 min, why didn't anybody tell me about this thing before ?!?
6
7
45
@pmelsted
Pall Melsted
8 years
5yo daughter: "we really only have 4 fingers, see 0,1,2,3,4" #soproud #csparenting
2
16
44
@pmelsted
Pall Melsted
8 years
Whoever decided to name the yeast chromosomes I-XVI rather than 1-16 was not a fan of shell scripts
3
26
43
@pmelsted
Pall Melsted
8 years
Next time I give a talk in England this will be my title slide
Tweet media one
2
8
43
@pmelsted
Pall Melsted
8 years
Don't get me wrong, I love R, some of my best friends use R, but it's the javascript of data science (not a complement)
2
27
40
@pmelsted
Pall Melsted
4 years
@mbeisen If you are using R for data analysis just drink the tidyverse koolaid. This is a very helpful book and how I got up to speed using it
1
2
39
@pmelsted
Pall Melsted
1 year
Suffix tree / k-mer / FM-index
Tweet media one
1
3
37
@pmelsted
Pall Melsted
7 years
tail -n+2 removes the first line from the input #whywasntItaughtthis
7
6
37
@pmelsted
Pall Melsted
5 years
Thesis defense starting for @hannespetur my first PhD student, joint with @bvhalldorsson
Tweet media one
1
3
37
@pmelsted
Pall Melsted
7 years
There is p-value hacking and then there is p-value hammering
1
7
37
@pmelsted
Pall Melsted
1 year
22 years of studying sorting algorithms finally paid off
Tweet media one
3
2
36
@pmelsted
Pall Melsted
4 years
I just had a covid test as a random sample from the population. I timed the visit, 4 minutes and 50 seconds from entering the building until walking out again. Results are promised within 24 hours, but I’m guessing I’ll know later tonight.
2
5
34
@pmelsted
Pall Melsted
1 year
gzip is the new logistic regression
2
2
34
@pmelsted
Pall Melsted
4 years
The data behind this figure is based on contact tracing for about 1200 individuals. Each trace requires about 20-200 phone calls and is done by a dedicated team at @almannavarnir working hard to contain the spread
@antonioregalado
Antonio Regalado
4 years
The report from DeCode and Kari Stefansson is out. "Spread of SARS-CoV-2 in the Icelandic Population" They tested 6 % of Iceland population. Found 0.6-0.8% infected. Charts show transition from imported case to family spread in a month.
Tweet media one
4
62
80
0
8
33
@pmelsted
Pall Melsted
7 years
It's not just that I hate the GTF format, I hate that nobody can agree exactly on how to use it #ivegot99bioinfoproblemsandGTFisallofthem
2
6
33
@pmelsted
Pall Melsted
3 years
A member of parliament in Iceland(stepping in as a substitute for another MP who is infected with Covid) is trying to get an isolation of Covid positive individuals thrown out in the courts. Ironically, not for the MP he is substituting for. 1/4
2
6
33
@pmelsted
Pall Melsted
5 years
If you’re feeling extra charitable you can stop calling them a geek
@rafalab
Rafael Irizarry
5 years
Dear everybody, If you have to choose one nice thing to do for the computer geek helping you, don't use spaces in your filenames. Instead of "My Document", use "my-document", "my_document" or "myDocument" Spaces indicate the end of the filename in some of the tools we use.
17
196
736
3
6
32
@pmelsted
Pall Melsted
3 years
0.4% of all Icelanders were diagnosed PCR positive for Covid yesterday. Scaled to the US that’s like 1.5M in one day. Elementary schools will open Jan 4th and 5-11y vaccinations will begin on Jan 10th. What could possibly go wrong.
3
3
31
@pmelsted
Pall Melsted
1 year
@iddux Bioinformatics algorithms by Pevzner and Compeau. The alignment chapter alone is worth it because it emphasizes understanding over implementation details. Rosalind problems are a fantastic addition to just studying the text
4
6
31
@pmelsted
Pall Melsted
4 years
Contact tracing has significantly reduced the spread of the virus. But it’s not enough by itself. Do your part. Limit social interactions, wear a mask, get tested ASAP for symptoms /end
3
2
32
@pmelsted
Pall Melsted
7 years
If you feel like there is an impeding doom where deep learning will eat your data analysis lunch, just watch this
@BeEngelhardt
Barbara Engelhardt
7 years
My TEDx talk on the importance of building interpretable, open box machine learning models and using domain-informed detective work to accelerate discovery in genomics, biology, and medicine:
Tweet media one
6
188
511
1
6
31
@pmelsted
Pall Melsted
6 months
We argue that it is better to not modify the cell-by-gene matrix at all but rather to correct objects which affect clustering. Downstream statistical tests can then take the batch identifier as a covariate, e.g. linear models or MAST. 6/N
1
3
31
@pmelsted
Pall Melsted
9 years
Introducing the Bioinformatics Ironman: write an assembler, a short/long read aligner and a file format
3
17
30
@pmelsted
Pall Melsted
4 years
Very interesting analysis of the choices made when designing the Biontech/Pfizer mRNA vaccine. So much of this builds on top of decades of basic scientific research. Our public research funds hard at work.
0
9
29
@pmelsted
Pall Melsted
3 years
Omicron’s fake move of (maybe) appearing to be a milder disease while spreading faster is an ankle-breaker and politicians seem to be falling for it.
1
7
29
@pmelsted
Pall Melsted
7 years
Bioinformatics pro tip: keep an unpatched system to benchmark your programs and a patched one for anybody else's
2
12
27
@pmelsted
Pall Melsted
6 years
After one week examine the tooth and discuss what happened and the likely causes. Ask them to write up a short report 2/n
1
6
28
@pmelsted
Pall Melsted
7 years
Graphtyper paper is out congrats to my phd student @hannespetur and co: @bvhalldorsson @hakon_jon @BirteKehr
1
15
27
@pmelsted
Pall Melsted
2 years
Open up Rstudio library(writexl) ?write_xlsx wait for it ...
2
5
28
@pmelsted
Pall Melsted
2 years
Great start for Genome Informatics 2022, first in-person meeting since 2019, with @ceclindgren giving the first keynote #gi2022
0
3
27
@pmelsted
Pall Melsted
3 years
Vaccinated
Tweet media one
Tweet media two
1
0
26
@pmelsted
Pall Melsted
7 years
Graphtyper, first paper/preprint by my phd student @hannespetur is out @bvhalldorsson @decodegenetics
0
27
27
@pmelsted
Pall Melsted
9 months
That's a wrap for a great Genome Informatics conference #gi2023 See you next year in Hinxton, UK btw, the cover had a small easter egg, let me know if you can find it
Tweet media one
1
4
26
@pmelsted
Pall Melsted
10 years
Tweet media one
4
29
27
@pmelsted
Pall Melsted
6 years
. @ggonnella talking about GFA format , grew out of a blog post …, @lh3lh3 proposed GFA in another blog post …, specification came later and improved to GFA2 which generalized better for long read technologies #GI2018
1
15
25
@pmelsted
Pall Melsted
4 years
Important life lesson: geologist use the concept “it will happen in the next hours or so” much like bioinformaticians
1
5
27
@pmelsted
Pall Melsted
6 years
De Bruijn graphs with even k are just fine, also it doesn’t matter how you pronounce de Bruijn
0
2
27
@pmelsted
Pall Melsted
3 years
Mix-n-match second dose today: Jansen + Moderna, 7 weeks apart. Will report on side effects (n=1)
Tweet media one
2
0
27
@pmelsted
Pall Melsted
3 years
Tweet media one
0
2
25
@pmelsted
Pall Melsted
5 years
8yo daughter: Hey Siri can you do my homework? Siri: in the words of Aristotle “the roots of education are bitter, but the fruit is sweet”
0
5
27
@pmelsted
Pall Melsted
9 months
View from the balcony
Tweet media one
2
3
26
@pmelsted
Pall Melsted
9 years
I'm surprised it took me this long to figure out what you call it when you use someone's data: Paracitation #researchparasites
1
23
26
@pmelsted
Pall Melsted
6 years
I’m excited about this and will be until my first poster segfaults
@NatalieTelis
Natalie Telis
6 years
This is really awesome! Next step is to just display a Shiny app where people can play with the data, and really interact with the story - not frozen figure images. The future looks cool 😎📊📈
2
2
19
0
1
26
@pmelsted
Pall Melsted
2 years
These UMAP jokes have gone too far
Tweet media one
0
4
25
@pmelsted
Pall Melsted
3 years
I'll give a short talk on the value of sequencing all SARS-CoV-2 samples in Iceland as part of the #COVID19Nordic research response workshop This Thurs with free reg Great talks from all the Nordics detailing strategies for dealing with the epidemic
Tweet media one
1
5
26
@pmelsted
Pall Melsted
4 years
@lpachter This is an improvement, the previous version had a fifth degree polynomial
Tweet media one
1
1
24
@pmelsted
Pall Melsted
4 years
Polar stratospheric clouds
Tweet media one
0
4
25
@pmelsted
Pall Melsted
7 years
Fasta has no spec so it's just cool if I use :Þ as a field separator in the name, right?
4
3
25
@pmelsted
Pall Melsted
9 years
BamHash, like md5sum but for comparing FASTQ to BAM: http://t.co/6GCAAnwsAo
2
24
25
@pmelsted
Pall Melsted
3 years
@michaelhoffman @bldgblog @metricausa @ianholmes The review for this paper should have been one of these bicycle memes
Tweet media one
0
1
24
@pmelsted
Pall Melsted
3 years
Two volcanoes, one dormant for 1800 years and the other ongoing, visible from our balcony
Tweet media one
Tweet media two
0
2
24
@pmelsted
Pall Melsted
9 years
High Speed Hashing for Integers and Strings http://t.co/qhkqa9Q2Rx by Mikkel Thorup. Well worth reading for hashing afficionados
4
17
22
@pmelsted
Pall Melsted
7 years
This happened one year ago today and yes I followed through on this
@pmelsted
Pall Melsted
8 years
Next time I give a talk in England this will be my title slide
Tweet media one
2
8
43
0
1
23
@pmelsted
Pall Melsted
4 years
That’s some serious math. Your move Stanford
Tweet media one
2
1
22
@pmelsted
Pall Melsted
4 years
Yes we’re an island. Yes we are tiny. Reykjavík is not as dense as New York (but similar to urban US cities, eg Pittsburgh). Is it harder to scale this up? Yes, but totally worth it even if it is not 100% effective. #TestTraceIsolate + quarantine close contacts.
1
1
23
@pmelsted
Pall Melsted
3 years
I got my 3rd dose of vaccine (j&j, Moderna 100, Moderna 50) this morning. I can feel the aches starting up so now is a good time to grade that final exam.
1
0
23
@pmelsted
Pall Melsted
7 years
Listening to the sweet sound of fireworks after Iceland qualified for the World Cup
1
0
23