I am a postdoc researcher in computer science. I have a strong background in machine learning, and I am really fascinated by science in general. I would like to apply my knowledge to DNA related tasks, such as DNA sequencing/prediction, sequence assembly, or mutations/anomaly detection. So my question is: where do I have to start studying? Can I start directly with bioinformatics topics or do I need a strong genomics background? Can you suggest me resources of any kind (books, websites, etc.) to start learning this beautiful things?
What field of bioinformatics do you want to focus? There is the genomics but also proteomics and structural proteins, networks and many more (phylogeny, agricltural, non-human research...). Perhaps the easiest advice is to use some book as https://www.biostarhandbook.com/ to start (but is heavily focused in NGS),
How about reading some papers on that topic then and try reproducing/extending existing methods? Maybe you'll eventually be inspired to come up with a better way to do things and publish a paper of your own!
You can find dozens of papers by typing the words: machine learning sequencing, in google scholar.
Also, I'd argue against necessarily needing biology knowledge -- if you want to do methods development (e.g. come up with a better assembler or develop statistical methods or spectral clustering methods for cell x gene matrices in single cell genomics), you don't need much biology at all.
I've spent many years doing hypothesis-driven biology work and have authored papers in the field (which certainly required a solid biology background), but my current work is methods development, statistics, algorithms, data structures, raw sequencing data preprocessing, etc. which require very minimal biology knowledge. Knowing about ATP, the Krebs cycle, lysosome trafficking, etc. is completely unnecessary for my current work.
If you do want to answer biological questions (e.g. using NGS to discover which genetic adaptations cause drug-resistant cancer cells to arise), then biology knowledge is essential.
Biology. Get domain knoelwdge. It is very common that experts from other fields try to enter life sciences with no background, and often that fails as results make no biological sense. Domain knowledge is key, and be it with "Molecular Biology Of The Cell", a standard textbook.
Thank you. Since I thought about having a biology background, I ordered the book 'Lewin's Essential Genes" (since I want to focus on dna sequencing and ngs). Is it a good book?
Possibly but this book may be more detailed than you would need to understand the concepts. Books are only going to take you so far. You should start working/collaborating with experimental scientists nearby. You will pick more things up by direct interactions than by books alone.
There is a surprising amount of decent introductory bioinformatics on Youtube in the meantime.
Also, Galaxy training network provides good explanations of many of the x-seq NGS approaches, such as RNA-seq, ChIP-seq etc. There are also exercises which are probably invaluable.
nf-core provides very useful best practice workflows in nextflow, so you can get used to the tools.
I saw some introductory course on coursera which is one hour work and includes processing NSG data, blast and Galaxy formats. I am interested too and have PhD in Biology and Advanced Python and Machine learning but not genomics as such. But a one day training should get you started. However, knowing a bit of evolution and its forces, mutation and some genetics would help. But for machine learning you probably want gene discovery. Bioinformatics specialization course from University of San Diego would be a very fundamental course. I took it about 10 years ago and it was heavy for me at the time due to its Python and computer theories but it should be much easier now after having some computer coding skills on string manipulation. I highly recommend this course and its Book as well Phillip I guess, a Russian.
I hope to get into gene prediction myself.
What field of bioinformatics do you want to focus? There is the genomics but also proteomics and structural proteins, networks and many more (phylogeny, agricltural, non-human research...). Perhaps the easiest advice is to use some book as https://www.biostarhandbook.com/ to start (but is heavily focused in NGS),
I'd like dna sequencing and ngs, if it is possible to use machine learning approaches with those topics!
How about reading some papers on that topic then and try reproducing/extending existing methods? Maybe you'll eventually be inspired to come up with a better way to do things and publish a paper of your own!
You can find dozens of papers by typing the words: machine learning sequencing, in google scholar.
Also, I'd argue against necessarily needing biology knowledge -- if you want to do methods development (e.g. come up with a better assembler or develop statistical methods or spectral clustering methods for cell x gene matrices in single cell genomics), you don't need much biology at all.
I've spent many years doing hypothesis-driven biology work and have authored papers in the field (which certainly required a solid biology background), but my current work is methods development, statistics, algorithms, data structures, raw sequencing data preprocessing, etc. which require very minimal biology knowledge. Knowing about ATP, the Krebs cycle, lysosome trafficking, etc. is completely unnecessary for my current work.
If you do want to answer biological questions (e.g. using NGS to discover which genetic adaptations cause drug-resistant cancer cells to arise), then biology knowledge is essential.
It really depends on what you want to do.