Are postdoc job ads these days asking too much from a PhD graduate?

[Melissa]

Looking at the bioinformatics postdoc job ads these days, I can't help but wonder if..
a) Some job ads are asking too much from a fresh phD graduate?
a) Or are they trying to recruit people with postdoc experience? A fresh PhD graduate will not stand a chance
c) Do they really find someone who fulfill all the criteria?

The reason I'm saying this is I often find postdoc job ads list many requirements. The common requirement is NGS data analysis but it's not just any NGS data analysis! It has be a combination of 3-5 skills such as... *Ahem clear my throat... genome assembly, transcriptome sequencing, gene expression, Chip-Seq, miRNA sequencing, methylation study etc. In addition, one must have experience in Unix environment and scripting (Perl || Python) && R && Awk && (C++ || Java) && *SQL*. On top of that, candidates have to be highly motivated, have nice publications, work in a collaborative environment, have good communication skills (See Appendix I).

I once looked at a postdoc bioinformatics job ad with my colleague who is an experienced bioinformatician. We only fulfill only 7-8 out of 10 requirements. I mentioned this to a senior researcher. He fulfills all the requirements but the job can't afford to pay him. He thinks that what is most important is the ability of the candidate to learn any skill quickly.

What is your take on this?

[vivek_]

In my opinion post doc salaries are absurdly low for the level of talent and expertise they provide/are expected to provide.

[Melissa]

That's true. But this is the only way to start in academia. Postdoc positions have become so competitive because there are not many jobs around.

[Brian Bushnell]

Job postings routinely ask for impossible sets of qualifications; I wouldn't worry about it. They have to choose from people that apply, and nobody who fulfills absurd requirements will apply for a job that pays very little.

[Melissa]

You have a point. Occasionally you do get people who are over-qualified. Recently, my lab received an application for a lab technician position from somebody who has a PhD and is interested to move here.

[gringer]

"We require 5 years experience with the Nanopore MinION!"

I don't think past experience is necessary for job requirements. Whatever the job, you're still going to need to spend a few weeks (or months) training new people up. People who can learn how to do things quickly are much better than people who know how to do lots of things (although those two frequently go together). Productivity is lost while you are training a new person for the job, so the quicker you can get out of that rut, the better.

I would prefer a person who has a wide range of skills (but not including everything in the job application) to a person has only experience in the specific categories that are on the job application (regardless of how good they are at those skills).

[Wallysb01]

What Brian said. Job ads are wish lists. Often times when I’m looking at job ads I see combinations of skills that are next to impossible for anyone at that experience level to have.

I love seeing all the jobs asking for CRISPR experience right now, for example. That technology has been around long enough for how many people to actually COMPLETE a project using it? I see 225 hits in pubmed when searching "crispr cas9”, which probably means maybe 500 people have really participated in a significant role in a published crispr/cas9 paper. Now how many of those people are even on the job market at a given experience level?

So, even if you don’t hit what appears to be a major job qualification, don’t be deterred from putting in an application. Most employers quite frankly seem very out of touch with what actually exists out there or they know they won’t get everything they want already.

[dpryan]

I'll echo the "what Brian said" sentiment. I get the feeling that many ads are written by HR departments or PIs who don't actually have a clue.

[GenoMax]

Unfortunately the applications may be (pre)screened by someone in HR who would not necessarily have subject matter expertise needed. They may be following a checklist (hopefully created by someone in the know) and if the application doesn't don't happen to "hit" x out of y minimum, it may die a "premature" death right there before ever reaching the eyes of a hiring manager.

I see that this thread was posted in the "academic" jobs section so this may not always apply (one would like to think).

[rskr]

I've noticed a couple trends. One is kind of like multi skilled players trend in basketball, shooting centers, point forwards, combo guards, swing men, etc. as has has been noted by sports commentators there are more players with more skills these days, but the players aren't necessarily more skilled(though some like LeBron are). For some teams these players serve some strategic purpose, and some times they just don't fit in the system.

Another trend I have noticed is that listings require many skills, but many of the skills are rather trivial. For example, Perl, python, Java, even R are pretty much just different ways of calling C++ libraries, job specific lingos and data structures that can easily be picked up if you have studied C++ in depth a Perl hash is a no brainer, though few jobs actually require c++.

Another, I have noticed is that often times employers are just trying to replace someone that left(for more money) who happened to use all of those technologies.

Another, is that previous generations of techies got there jobs with relatively few skills(they knew Perl and maybe HTML), but schools have ramped up on producing bioinformaticians with these skills.

[gringer]

For example, Perl, python, Java, even R are pretty much just different ways of calling C++ libraries

Using these programming languages properly for more than a basic use requires a bit more knowledge. In particular, R has a lot of functional elements to it that trip up people trying to do advanced programming -- imperative programmers will tend towards multiple 'for' loops instead of using apply, aggregate, etc., and don't realise that many functions are trivially vectorisable.

[rskr]

Using these programming languages properly for more than a basic use requires a bit more knowledge. In particular, R has a lot of functional elements to it that trip up people trying to do advanced programming -- imperative programmers will tend towards multiple 'for' loops instead of using apply, aggregate, etc., and don't realise that many functions are trivially vectorisable.

Sure, but how many times have you read the library it was calling into because the feature wasn't documented well enough in the cheesy vignette? A great example is how does the corr function handle missing values with certain parameters, you have to read the C to find out. Besides, the C++ parser and compiler is a lambda language itself(boost is built on this idea), so if you really know C++ you also know functional concepts like lambda, lists and recursion.

[Brian Bushnell]

Java is actually not a way of calling C++ libraries. Calling C/C++ code from Java can sometimes increase performance, but it's non-trivial and gives up many of Java's advantages; furthermore, most library classes are implemented in pure Java. So it does not happen much in practice (aside from low-level, OS-specific code).

In my experience, most Python/Perl code is written to either perform tasks directly (even computationally-heavy ones, regardless of whether it's a good idea), or invoke other standalone programs in a pipeline; using them to call C++ library methods seems to be fairly rare in bioinformatics. Many people who prefer to program in Python and Perl are less likely to be familiar with C/C++ and related libraries, or prefer portability over performance. Often, programs are written for tasks that are expected to be one-time but unexpectedly become common. And multi-language chimeric programs tend to be hard to install.

As for parsers, lexers, and compilers... they can be written in anything, and most code is written by people with no idea of how they work. Functional languages may work better for parsers and lexers, but I don't see any particular advantages for compilers or assemblers.

One of the advantages of Java is the excellent and consistent documentation of library classes. I have never looked at the source code to discern how a library function works (other than for performance or curiosity) - I just use it according to the description, and if it is too slow, I re-implement it myself. I have yet to encounter a case when the documentation was insufficient or unclear regarding functionality.

To return to the topic, I think it's sufficient to ask for "programming experience" for bioinformatics. The assumption should be that an intelligent person can quickly obtain basic knowledge of any mainstream programming language, if they can already program in any language. Unless the position is specifically for software development, the language is immaterial. If the intention is to develop robust, high-performance software as part of a team, then a single language should be specified in the job requirements, and the opening should specifically seek programmers, not bioinformaticians. Job openings that list multiple languages (unless they are assembly languages) indicate that there is actually no specific language requirement, just knowledge of programming.

Sequencing-platform-specific knowledge (on the digital side) can be obtained in weeks. There is no good reason to require any new hire to have any knowledge of a specific platform unless you do not have anyone with that knowledge, and capable of training others.

All I care about (from a new employee) is creativity, ability to learn, ability to communicate, intellectual curiosity, flexibility, honesty, and friendliness. Most of these cannot be evaluated fairly by checkpoints on a list, so I see HR mainly as a barrier to qualified people getting useful jobs. A good set of job requirements might be:

"Graduated from high school. Friendly. Capable of holding a conversation. Shows interest in area of job duties. Does not blatantly lie during interviews."

To avoid false negatives, anyone who fulfills those needs further review by someone conversant in the area of interest.

[rskr]

Java is actually not a way of calling C++ libraries. Calling C/C++ code from Java can sometimes increase performance, but it's non-trivial and gives up many of Java's advantages; furthermore, most library classes are implemented in pure Java. So it does not happen much in practice (aside from low-level, OS-specific code).

In my experience, most Python/Perl code is written to either perform tasks directly (even computationally-heavy ones, regardless of whether it's a good idea), or invoke other standalone programs in a pipeline; using them to call C++ library methods seems to be fairly rare in bioinformatics. Many people who prefer to program in Python and Perl are less likely to be familiar with C/C++ and related libraries, or prefer portability over performance. Often, programs are written for tasks that are expected to be one-time but unexpectedly become common. And multi-language chimeric programs tend to be hard to install.

As for parsers, lexers, and compilers... they can be written in anything, and most code is written by people with no idea of how they work. Functional languages may work better for parsers and lexers, but I don't see any particular advantages for compilers or assemblers.

One of the advantages of Java is the excellent and consistent documentation of library classes. I have never looked at the source code to discern how a library function works (other than for performance or curiosity) - I just use it according to the description, and if it is too slow, I re-implement it myself. I have yet to encounter a case when the documentation was insufficient or unclear regarding functionality.

To return to the topic, I think it's sufficient to ask for "programming experience" for bioinformatics. The assumption should be that an intelligent person can quickly obtain basic knowledge of any mainstream programming language, if they can already program in any language. Unless the position is specifically for software development, the language is immaterial. If the intention is to develop robust, high-performance software as part of a team, then a single language should be specified in the job requirements, and the opening should specifically seek programmers, not bioinformaticians. Job openings that list multiple languages (unless they are assembly languages) indicate that there is actually no specific language requirement, just knowledge of programming.

Sequencing-platform-specific knowledge (on the digital side) can be obtained in weeks. There is no good reason to require any new hire to have any knowledge of a specific platform unless you do not have anyone with that knowledge, and capable of training others.

All I care about (from a new employee) is creativity, ability to learn, ability to communicate, intellectual curiosity, flexibility, honesty, and friendliness. Most of these cannot be evaluated fairly by checkpoints on a list, so I see HR mainly as a barrier to qualified people getting useful jobs. A good set of job requirements might be:

"Graduated from high school. Friendly. Capable of holding a conversation. Shows interest in area of job duties. Does not blatantly lie during interviews."

To avoid false negatives, anyone who fulfills those needs further review by someone conversant in the area of interest.

Please go look up GTK then communicate to the world your ability to learn a how java uses a C library to do graphics.

[Brian Bushnell]

Graphics and I/O are OS-specific.