Sanju Sinha
@Sanjusinha7
Followers
3,890
Following
197
Media
252
Statuses
2,681
Asst Prof @sbpdiscovery I tweet comp bio-resources. | PhD in ML→cancer therapies| ex-NCI, UMD, IITG, Max Delbruck, NCL| Tweets are chatGPT's view, not mine.
San Diego, California
Joined June 2012
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
and more
• 1792303 Tweets
LINGORM PANTENE TRIP
• 639212 Tweets
Bruno Mars
• 501282 Tweets
#ROSÉ_BRUNO_APT
• 235061 Tweets
Al Smith
• 142124 Tweets
APT OUT NOW
• 136472 Tweets
Corinthians
• 131385 Tweets
Flamengo
• 131332 Tweets
Saints
• 91448 Tweets
Yankees
• 90801 Tweets
#Happy20thFOTFOT
• 81589 Tweets
Schumer
• 72088 Tweets
CONGRATS CL IG1M400K
• 71242 Tweets
Broncos
• 44092 Tweets
Cleveland
• 40911 Tweets
Allan
• 35959 Tweets
#NuNewJapanDebut
• 30161 Tweets
対象作品
• 23156 Tweets
San Marcos
• 22766 Tweets
Patrick New Single
• 21599 Tweets
NUNEW DEBUT IN JAPAN
• 19700 Tweets
超電ブレイカー
• 19272 Tweets
Ganso
• 19232 Tweets
David Fry
• 18573 Tweets
Big Christmas
• 15693 Tweets
Mookie
• 15641 Tweets
Bo Nix
• 15591 Tweets
Lakers
• 14849 Tweets
Rattler
• 13824 Tweets
集英社秋マンガチャ開催
• 12094 Tweets
ミミッキュ
• 11762 Tweets
Dennis Allen
• 11495 Tweets
Kamara
• 10342 Tweets
フラワー
• 10061 Tweets
Pinned Tweet
With a mixed bag of emotions - nervousness, excitement, fear & determination, I am announcing the start of my lab at
@sbpdiscovery
, San Diego.
Our focus will be to develop new preventative therapies for cancer using machine learning.
19
24
112
A lot of Machine Learning (ML) I learned during my Ph.D. was from youtube. I didn't have a guide to do this effectively and thus here it is:
A complete guide to studying ML from youtube: 13 best and most recent ML courses available on YouTube. 👩🏫🧵⤵️
210
2K
11K
Our understanding of the immune system is quickly growing.
11 resources (videos and papers) covering the fundamentals and computational tools available to study the immune system. 🧵👇🔬🤒
11
261
1K
Have you wondered how Yamanaka found the 4-gene combination to transform somatic cells to stem cells? The # of possibilities are greater than - 20,000^4.
I did this yesterday & read the original paper.
What a journey & pure joy it was! Here’s how he did it. ⌛️
6
139
765
@OpenAI
Folks! What about a search bar to find a specific chat from your chat history to use already provided info in prev sessions? Simple and can be so helpful. C’mon!
46
8
506
Wow! This is the most well-curated and relevant bioinfo around course covering standard bioinfo to utilizing scRNA-seq, genomics, precision oncology, immunotherapy, CRISPR screens (Chp 19-26).
@XShirleyLiu
@joshuastarmer
@tangming2005
@getz_lab
@twang5
3
117
434
We will start with "Stanford CS229: Machine Learning" by Andrew Ng to start and learn the following ML concepts:
Linear & Logistic Regression,
Naive Bayes, SVMs, Kernels
Decision Trees, Introduction to Neural Networks
Debugging ML Models.
6
49
403
I am hiring multiple Computational Biologists for my new lab
@sbpdiscovery
, San Diego.
An alternative to a postdoc, it will pay higher (70-90K$; Be the change you want to see in the world) & is tailored to develop skills to industry transition.
My lab
13
165
408
Interested in aging and cancer. I did a year of literature survey on this.
Here is my list of 20 key open questions and challenges to better understand the interplay between aging and cancer. A thread 🧵👇
3
81
378
Tumor transformation due to inflammation is an emerging field.
The claim that inflammation plays a major role in carcinogenesis has been there for a while. However, a shift in our understanding happened when a few years ago deep sequencing of aged normal tissues revealed that
6
51
298
Best computational practices to analyze Spatial Transcriptomics (ST) are yet non-trivial.
14 Resources, including videos, papers, data repo, tutorial, & a podcast, covering our current understanding of preprocessing & downstream analysis of Spatial Transcriptomics. 🌌🧬 🧵👇
3
74
286
First demonstration of AlphaFold application to generate a novel drug from scratch to target a gene of interest.
This can be a game changer for early drug discovery.⚡️
2
56
255
Beyond an AI genius, Andrej Karpathy is a brilliant teacher. His creative teaching methods make this intro to Neural Networks (NN): Zero to Hero makes one of the best ways to get introduced to NN.
1
23
207
A series of mini-lectures (~5 mins) covering various introductory topics in ML by Cassie Kozyrkov, covering:
Explainability in AI, Precession vs. Recall, Statistical Significance, Clustering and K-means, and finally, Ensemble models.
1
15
202
Spending millions of $, Grail created a TCGA-like study to systematically answer the following:
What is the best cell-free DNA method to detect cancer from blood?
17 things we learned.🧵🩸
3
36
171
"MIT: Deep Learning for Art, Aesthetics, and Creativity " covers the application of deep learning for art, aesthetics, and creativity, including Neural Abstractions, Efficient GANs, and explorations in AI for Creativity.
1
11
148
(Striking again) Andrew Ng's "Stanford CS230: Deep Learning (2018)" covers:
The foundations of deep learning, how to build different neural networks (CNNs, RNNs, LSTMs, etc...), how to lead machine learning projects and finally - AI and Healthcare.
1
10
131
Applied Machine Learning teaches some of the most widely used techniques in ML, including:
Optimization and Calculus, Overfitting and Underfitting, Regularization, Monte Carlo Estimation, and Maximum Likelihood Learning.
2
12
122
@pdhsu
Not an extension but
Spellbook -
Legalobot -
are working on these lines. The second is LLM-based, like chatGPT. Both have a waiting list to use, unfortunately.
1
14
110
The first part of 'Practical Deep Learning for Coders' teaches you how to:
Build & deploy deep learning models for vision & NLP. Use PyTorch, plus popular libraries like fastai.
1
9
108
Methods & data available to you are your thinking tools. While I learned the methods in my classes, I wish I knew various data available to me.
10 resources to learn almost all the big data resources available in cancer research. 🧵👇
5
25
106
To folks applying to grad school from developing countries & see the crazy application fee.
See this list of grad programs in US with application fee completely waived off: (by
@cientificolatin
)
Also, attend the Open house (virtual) of the programs you
2
59
105
This course focuses on the probability and maths behind ML, covering:
Reasoning about uncertainty, Continuous Variables, Sampling, and Markov Chain Monte Carlo.
1
8
94
The 2-hour second part of 'Practical Deep Learning for Coders' takes a deep dive into a recent hot ML topic - Diffusion Models.
3
5
92
ML with graphs teaches some of the latest graph techniques in machine learning:
PageRank, Matrix Factorizing, Node Embeddings, Graph Neural Networks, Knowledge Graphs, and finally, Deep Generative Models for Graphs.
1
11
85
8 out of 10 ML breakthroughs you recently heard of are likely based on transformers. "Stanford CS25 - Transformers United" aims to introduce us to the following:
Transformers, its applications in Language (GPT-3), vision & universal compute
engines.
1
9
77
'Foundation Models' is a recent course (June 2022) that aims to teach about foundation models like GPT-3, CLIP, Flamingo, and cross-language generalization.
1
5
76
Using big data in healthcare. Here are 10 educational resources for anyone interested in building skills to analyze big data in healthcare.
Ranging from introductory to advanced, this includes courses, youtube channels, papers & online books.🧵🥑👇
2
19
75
This 12-part Deep Unsupervised Learning aims to teach the latest and most widely used techniques in deep unsupervised learning:
Autoregressive Models, Latent Variable Models, & Self-supervised learning.
1
6
70
I curated a list of 28 common issues one faces while using machine learning for biomedicine research and using different kinds of omics data. I also provided guides on how to best overcome them. 🍉🧵👇
4
20
61
This exercise certainly taught me that reading original nobel works and summarizing it is a task of pure joy. :)
Here’s the original paper. Cheers!
9
3
60
An under-the-radar🚨 paper from two weeks ago answered:
Why certain pateints might be more vulnerable to adverse effects during immunotherapy? & Can we identify them before or during therapy? 🧵🛠️
1
16
60
There have been 100s of exciting biomedicine discoveries in the last year.
Here are my top 11 biomedicine discoveries primarily using big data in 2022 that will likely have a long-lasting & most significant impact.🧪🫑🧵
2
18
58
How is Tumor Mutation Burden related to ICB response? Our understanding of this relationship is constantly updating and improving as we get more data.
Here are 11 recent findings about this relationship.🧵
1
14
56
We are not done with single cell.
With 190 unique scRNA-seq datasets from 20 diff tumor types, TISCH (Tumor Immune Single-cell Hub) is a portal to download their *uniformly* processed read counts & their metadata. Go visualize your biomarkers/genes here.
1
13
48
Reading two technical but philosphically sharp papers for today (Insights soon):
Pre-cancerous Niche Remodelling Dictates Nascent Tumour Survival
Cellular adaptation to cancer therapy along a resistance continuum
0
3
42
20 open grand challenges to understand the relationship btw cancer and aging better.
Interested in aging and cancer. I did a year of literature survey on this.
Here is my list of 20 key open questions and challenges to better understand the interplay between aging and cancer. A thread 🧵👇
3
81
378
8
4
39
I am developing a drug discovery startup based on a recent computational technology I co-developed (See below).
I would love to chat if you would like to collaborate on this or a potential investor. DM/email sanju
@terpmail
.umd.edu
Drug target identification is at the heart of drug development, and we’ve been working to change how it’s been done.
We present DeepTarget: a new computational tool to characterize a drug’s mechanism of action in-depth beyond its primary target. 🧰🧵👇
6
95
332
17
6
38
A single measure of tumor-specific total mRNA expression from bulk sequencing using deconvolution predicts an increased risk of disease progression and death.
1
10
38
@tangming2005
Here’s my list of publicly available scRNA-seq profiles from patient tumors with clinical annotations. These patients are either treated with Chemo, Targeted or immunotherapy. Updated a couple of months ago & still building. Feel free to add.
3
2
37
A conceptual framework to examine the patterns of copy number changes in human cancer applicable WGS, WES, bisulfite sequencing, scDNA-seq & SNP6 revealing 21 signatures that explain the copy number patterns of 97% of samples.
1
20
36
Unsupervised discovery of tissue architecture in multiplexed imaging
This paper from 2022 is really creative - Finding tissue microstructures in an unsupervised manner.
0
7
34
Our institute is looking for a faculty in the field of Computational Biology and Artificial Intelligence.
HR didn't allow me to add this part to the job post, so here it is - "We have the best tacos and beaches in the country" ;)
0
17
36
I will add here as I find more.
I tweet resources for big data research in healthcare. Follow me
@Sanjusinha7
if that is of interest. See below for other such resources.
6
4
33
Coming to the real deal - The computational tools.⚡️
We will start with a ‘how-to guide’ for using omics for systems immunology and explain how current single-cell-level omics technologies can be applied to ask immunological questions.
1
5
32
Our results suggest that "stochastically accumulating changes in any set of data that have a ground state at age zero are sufficient for generating aging clocks."
tldr - There's a possibility aging clocks might be just noise.
2
2
31
Hey Cancer research twitter!
I am creating a single table comprising the big-data resources for cancer research. I think this could be a helpful resource. I wonder if you have a few minutes and could add the ones I am missing here. Thank you!
0
14
31
28 common issues you will likely face while using ML for biomedicine and how to address them.
I curated a list of 28 common issues one faces while using machine learning for biomedicine research and using different kinds of omics data. I also provided guides on how to best overcome them. 🍉🧵👇
4
20
61
1
4
30
The immune system in cancer.🦀
This study highlights *almost all* the computational tools for interrogating cancer immunity, discusses the advantages and limitations of the various methods, and provides guidelines to assist in method selection.
2
2
30
10 educational resources for anyone interested in building skills to analyze big data in healthcare.
Using big data in healthcare. Here are 10 educational resources for anyone interested in building skills to analyze big data in healthcare.
Ranging from introductory to advanced, this includes courses, youtube channels, papers & online books.🧵🥑👇
2
19
75
1
8
25
I am from the socio-economic background where my parents weren’t able to pay my high school tuitions on time. Today I have been awarded as one of the 13 “Emerging Leaders of Computation Oncology” in the world from
@CompOncMSK
5
1
26
Before we go on to learn the computational tools to quantify and measure the immune system, Let's start with a 15 mins visual overview of the fundamentals of the immune system.
1
3
27
A list of almost all the big data resources available in cancer research.
Methods & data available to you are your thinking tools. While I learned the methods in my classes, I wish I knew various data available to me.
10 resources to learn almost all the big data resources available in cancer research. 🧵👇
5
25
106
1
3
24
Elated and proud to present the most important discovery of my Ph.D. - We found that genetic editing using CRISPR-Cas9 selects for mutant forms of two cancer genes (Kras and p53) calling for cautious monitoring of their mutation status during CRISPR-Cas9 therapeutics.
1
2
25
11 computational resources to study immune system
Our understanding of the immune system is quickly growing.
11 resources (videos and papers) covering the fundamentals and computational tools available to study the immune system. 🧵👇🔬🤒
11
261
1K
1
2
24
A single blood measure - Ly6Ehi neutrophils, predict immunotherapy response in humans with high accuracy (average AUC ≈ 0.9) in multiple cancer across ~1500 patients.
If robust, this has high-TMB biomarker level potential. (We will know soon.)
0
6
23
In a long-awaited study testing molecular aging clocks longitudinally across 1K individual over 7 years, authors found *No* evidence for associations between change in various methylation clocks (Horvath’s, GrimAge) with physical & mental deterioration.
1
4
24
A lay summary of our paper presenting a proof of concept that single-cell transcriptomics can be used to guide treatments in clinics.
Check out below!
0
3
25
12 resources to best analyze Spatial Transcriptomics.
Best computational practices to analyze Spatial Transcriptomics (ST) are yet non-trivial.
14 Resources, including videos, papers, data repo, tutorial, & a podcast, covering our current understanding of preprocessing & downstream analysis of Spatial Transcriptomics. 🌌🧬 🧵👇
3
74
286
1
4
23
Going beyond matching genes only using their names.
0
4
24
Counting every win!
Grateful to inform that I have received the NCI Transition to Industry (T2I) Fellowship supporting advancement of my inventions at NCI towards a regulatory milestone, clinical trials, and subsequent commercialization.
3
2
24
Yamanaka and team won the nobel prize for showing that a defined combination of 4-factors can induce pluripotent stem cells from mere somatic fibroblast cells - a major leap forward in science.
2
0
22
Do you know that you can tell almost any drug’s target and its type by its name only?
Yes! The secret lies in the last characters of its name. Eg. In erlotinib - 'ib' refers to inhibitor, & 'n' denotes kinase | -mab stands for monoclonal antibody,
A map of these codes is below.
0
8
20
How was my experience starting a lab in 2023 against the backdrop of a world emerging from a pandemic?
@NatureCancer
gave me a platform to share my experience. Hope some of you find it insightful.
0
1
20
Cell-Cell interactions.
An overview of methods and tools to study Cell-Cell interactions from transcriptomic data and a highlight of the discoveries enabled by these analyses.
1
0
20
Proud to share that my recent work (
@NatureCancer
; ) done under the wonderful mentorship of
@bridmryan
@NCIEytanRuppin
was among the
@theNCI
CCR Milestones.
@NCIMedia
0
3
19
A novel anti-aging method in our T-cells.
Lanna & team revealed that T-cells acquire telomere from antigen-presenting cells upon contact with them & delay senescence and promote long-term immunological memory.
Nature is creative!
1
5
18
… 2) I think our current nomenclature of this stem cell business is inadequate and will likely be restructured - especially with emerging aging research. The cell transformation field is at its nascency.
Still, it is deserving of all accolades.
1
0
18
Examining these 10-factor pool & repeating the withdrawing one factor at a time among these 10 factors revealed 4 factors that were key.
They were Oct3/4, Sox2, c-Myc, and Klf4.
Again - these four factor were able to induce the needed colonies.
1
1
16
Machine learning (ML) is reimagining and redefining how we do healthcare research.
Here are the 13 most out-of-the-box and creative use of machine learning for healthcare research in 2022.🧄🧵
1
1
18
A thoughtful and beautiful review pointing complexities of cancer as a systemic disease - What a read!
Most importantly, it overcomes the reductionist nature of "hallmarks of cancer" concept.♥️
0
5
17
No two-gene combination was good enough to create colonies. While three-gene combinations (- Klf4/Oct4) were able to create small colonies, they were not able to maintain them.
Interestingly removing MYC changed the nuclear shape of the cells created. Not like usual stem cells.
1
0
15
So how did it go? The first experiment failed. No single factor was able to induce the stem cells. Not even a little - nada.
However, a combination of all 24 factors together showed 🥁🥁 22 resistant colonies - these are stem cells. So the solution seems to be there.
1
0
15
Please use the doc below to show your concern for a systematic effort against this.
0
1
17
When compared in terms of expression or morphologically, these stem cells were similar, but not identical to embryonic stem cells. (Big figures to show that)
1
0
15
This revealed 10 factors whose individual withdrawal resulted in no colony formation & fewer colonies after 16 days.
And then when these 10 were given in combination, they were good enough to produce the needed stem cell colonies. (We are getting closer.)
1
0
15
This is an exciting and brilliant tool to design small targeted panels that help you extract mutational signatures as reliably as the whole exome and has been shown to outperform MSK-IMPACT panels.
@lucmorrisnyc
@GeneCollector
@FoundationATCG
Check out our algorithm ScalpelSig, that designs genomic panels to best detect activity of mutational signatures. Presented at
#RECOMB2021
.
Led by Nick Franzese, in collaboration with Roded Sharan and
@maxleiserson
.
1
3
24
0
11
17
A series covering the update on TMB as immunotherapy (IT) biomarker & next challenges, a strategy to expand IT to hard-to-treat pediatric tumors, & how to target bystanders' T-cells and microbiome for IT. Upcoming pieces will focus on cancer vaccines etc.
1
6
17
@ElizSMcKenna
Hi! I am starting a lab tomorrow
@sbpdiscovery
, San Diego, focused on developing preventive therapies using machine learning.
I am nervous, excited, determined & afraid about this new step.
3
1
16
Other transcriptional factors tested are - Ecat1, Dppa2/3/4/5, Fbxo15, ERas, DNMT31, Ecat8, GDF3, SOX15, Fthl17, Sall4, Rex1, Utf1, Tcl1, *Klf4*, Beta-catenin, Grb2
However, even with 24, there are more than 330K possible combinations.🤹♂️
2
1
15
But first, how are we gonna really detect if something is a stem cell? They used a stem cell marker called “Fbx15”, & added a drug resistant gene with it before it.
So if cells survive under drug exposure, they are stem cells.
1
0
15
And we are done. Two thoughts:
1) Simultaneously, I think this paper is better than what I thought it would be and has more flaws than I would have thought.
This is so deeply satisfying and I am so happy about it. I do science to find moments like this. ..
1
0
16
Coming to Flaws: 1) The approach to remove one factor at a time is suboptimal. Will miss pairs that can buffer each other (OR gate). We also do not know how these four factors really induce this property (each of their job, steps, order, another nobel prize is here).
1
0
15
Yamanaka team drew the conclusion that: Oct3/4, Klf4, Sox2, and c-Myc play important roles in the generation of stem cells.
1
0
16
We actually did know that it is possible to induce stem cells from somatic cells. We also knew how.
How? A nuclear content transfer of somatic cells to an oocyte can create stem cells (same if they are fused with Embryonic cells). So what did they do?
1
0
16
@SusannaLHarris
I respects this side and this phenomenon fascinates me. I fail to wrap my head around it and would love to understand this. One of the best scientists of our era is religious and is doing more good than a lot of us. His book on this topic is in my list.
0
2
13
They chose 24 transcriptional factors - based on prior knowledge in stem cell research - Oct3/4 and c-MYC maintains pluripotent cells (& are also high in tumors - they did note this), SOX15/2 only upregulated in stem cells.
A major player favourite to win this was - 'Nanog'.
1
0
14
Methods and tools to study Cell-Cell interactions... (What again?)
We will study this in the context of cancer with a special focus on tumor-immune interactions.
1
1
15
Impetus grants from
@MartinBJensen
"We would rather fund the work you are most excited about doing, even if it might fail, than work that is certain to produce results but with limited impact on the field."
1
2
15
So we see no stem cells without Oct3/4 or Klf4, little without Sox2, without c-Myc cells with not really Stem-like - odd shape nucleus. Removal of the remaining factors had no effect.
1
0
14
Also, morphologically and in doubling time, these colonies were similar to stem cells. And now for the first time, we have defined-factors induced stem cells.
These cells were called iPS-MEF24 for “pluripotent stem cells induced from MEFs by 24 factors.”.
1
0
14
So which of the 24 candidates were really critical? You cannot test all combinations. Millions of combinations are possible.
How about we withdraw them one at a time? (smart, No?)
1
0
13
@NikoMcCarty
low-key thinking to do this for cancer data science. :) Wonder if you have to do it all over again, how would you do it this time?
1
0
14
Our recent work on CRISPR-risk was among the Top 25 most read articles of 2021 in
@NatureComms
(it was published in November ;) ). Let's celebrate victories, however small they are, during these insane times.
0
4
14
Next is another 15 mins visual. This explains the cells of the immune system and their different functions that provide an immune response to an invading pathogen.
1
1
14
🤯The metastatic spread of cancer that primarily occurs via the dissemination of circulating tumor cells (CTC) is concentrated & accelerated during sleep😴 due to high CTC generation dictated by key circadian rhythm hormones like melatonin & testosterone.
0
4
14
The creation of a synthetic minimal bacterium requires 149 proteins of unknown function. If they are important for a minimal cell, they are likely important for us. Understudied proteins deserves a hell of a lot attention than what they are getting now.
0
1
13
The authors introduce *Seq*, a high-performance, Python-based, compiled programming language geared toward biology that combines the ease of use of high-level languages like Python or Matlab with the runtime performance of low-level languages like C or C++.
A Python-based programming language for high-performance computational genomics
1
186
683
0
5
13
Next on our list is "Antigen."
What are the recent advances in predicting HLA binding and antigen presentation in human cells?
1
0
13